Completes the Windows-gating coverage for the built-in skills/ tree. Every
bundled SKILL.md now carries an explicit platforms: declaration so the
loader (agent.skill_utils.skill_matches_platform) can skip-load skills
that don't fit the current OS.
74 skills declared cross-platform (platforms: [linux, macos, windows]):
Creative (16): ascii-art, ascii-video, architecture-diagram, baoyu-comic,
baoyu-infographic, claude-design, creative-ideation, design-md,
excalidraw, humanizer, manim-video, p5js, pixel-art,
popular-web-designs, pretext, sketch, songwriting-and-ai-music,
touchdesigner-mcp
Autonomous agents: claude-code, codex, hermes-agent, opencode
Data/devops: jupyter-live-kernel, kanban-orchestrator, kanban-worker,
webhook-subscriptions, dogfood, codebase-inspection
GitHub: github-auth, github-code-review, github-issues,
github-pr-workflow, github-repo-management
Media: gif-search, heartmula, songsee, spotify, youtube-content
MCP / email / gaming / notes / smart-home: native-mcp, himalaya,
pokemon-player, obsidian, openhue
mlops (non-broken): weights-and-biases, huggingface-hub, llama-cpp,
outlines, segment-anything-model, dspy, trl-fine-tuning
Productivity: airtable, google-workspace, linear, maps, nano-pdf,
notion, ocr-and-documents, powerpoint
Red-teaming / research: godmode, arxiv, blogwatcher, llm-wiki,
polymarket
Software-dev: debugging-hermes-tui-commands, hermes-agent-skill-authoring,
node-inspect-debugger, plan, requesting-code-review, spike,
subagent-driven-development, systematic-debugging,
test-driven-development, writing-plans
Misc: yuanbao
5 skills gated from Windows (platforms: [linux, macos]):
mlops/inference/vllm (serving-llms-vllm)
vLLM is officially Linux-only; Windows requires WSL.
mlops/training/axolotl
Axolotl's flash-attn + deepspeed + bitsandbytes stack is Linux-first.
mlops/training/unsloth
Requires Triton + xformers + flash-attn — Linux only in practice.
mlops/models/audiocraft (audiocraft-audio-generation)
torchaudio ffmpeg backend + encodec dependencies are Linux-first.
mlops/inference/obliteratus
Research abliteration workflow; relies on Linux-focused pytorch
kernels and MLX — no first-class Windows path.
Same strict-over-lenient policy as the optional-skills sweep: when the
underlying tool's Windows support is rough, missing, or WSL-only, gate the
skill. Easier to un-gate after verified Windows support lands than to leak
partial support that manifests as mid-task failures.
Combined with prior commits in this branch, every bundled SKILL.md
(skills/ + optional-skills/) now has a platforms: declaration.
Hermes's skill loader (agent/skill_utils.skill_matches_platform) already honors
the 'platforms:' frontmatter field and skip-loads skills whose declared
platform list doesn't include sys.platform. Seven bundled skills are in fact
Linux/macOS-only but never declared it, so they leak into Windows skill
listings and sometimes load with broken instructions.
Audited all 160 SKILL.md files (skills/ + optional-skills/) for Windows-
hostile signals: apt-get/brew/systemd/chmod+x install flows, ptrace/proc
runtime dependencies, bash-only launcher scripts, and package dependencies
with no Windows build. The 7 below fail one or more of those tests in a way
that fundamentally can't be papered over by docs edits:
minecraft-modpack-server bash start.sh + chmod +x + apt openjdk
evaluating-llms-harness lm-eval-harness bash launcher scripts
distributed-llm-pretraining-
torchtitan bash multi-node torchrun launcher
python-debugpy remote attach relies on /proc ptrace_scope
pytorch-fsdp NCCL backend; Windows path is WSL only
tensorrt-llm NVIDIA TensorRT-LLM has no Windows build
searxng-search Docker volume flow assumes POSIX $(pwd)
All seven get 'platforms: [linux, macos]'. On Windows the loader now skips
them silently — no more phantom skill listings, no more mid-task failures
because an Apple-only path was surfaced as a suggestion.
Cross-platform skills that merely CONTAIN signals in examples or
install-instructions (brew install as one of several paths, /tmp/ in a code
snippet, etc.) are NOT touched by this commit. A broader audit that
declares the ~140 cross-platform skills as 'platforms: [linux, macos,
windows]' can follow as a separate change once each has been verified
working on Windows.
The installed user copies under ~/AppData/Local/hermes/skills/ (when they
exist) are also patched so the running session reflects the gating
immediately, but only the in-repo files are committed here.
Adds a dedicated '## Windows-Specific Quirks' section to the hermes-agent
skill so Windows pitfalls have one discoverable place to evolve. Inaugural
entries cover:
- Input / keybindings — Alt+Enter intercepted by Windows Terminal,
Ctrl+Enter as the Windows newline keystroke, mintty/git-bash behavior,
pointer to scripts/keystroke_diagnostic.py for investigation.
- Config / files — UTF-8 BOM HTTP-400 trap.
- execute_code / sandbox — WinError 10106 SYSTEMROOT root cause +
_WINDOWS_ESSENTIAL_ENV_VARS fix location.
- Testing / contributing — scripts/run_tests.sh POSIX-venv limitation and
the system-Python workaround, POSIX-only test skip-guard patterns.
- Path / filesystem — line-ending warnings (cosmetic), forward-slash
portability.
Collapses the old scattered Windows bullets under 'Platform-specific
issues' into a single pointer at the new dedicated section so there's
only one place to maintain this content.
Also adds the scripts/keystroke_diagnostic.py the skill now references —
a small prompt_toolkit Application that prints the Keys.* identifier and
raw escape bytes for every keystroke. Used to establish the Ctrl+Enter
= c-j fact on Windows Terminal; generally useful for anyone adding a
platform-aware keybinding.
Fifth and final slice polish on top of @dlkakbs's docs + skill. Three
things ship here:
1. Subscription renewal cron recipe (the #1 operational footgun).
Microsoft Graph webhook subscriptions expire at 72 hours max and
don't auto-renew. The shipped operator runbook mentioned
`maintain-subscriptions --dry-run` as a "daily or periodic check"
but never told operators how to actually automate it. Without a
scheduled job, any production deployment silently stops ingesting
meetings three days after go-live.
Adds an "Automating subscription renewal (REQUIRED for production)"
section to website/docs/guides/operate-teams-meeting-pipeline.md
with three concrete options and copy-pasteable configs:
- Option 1: Hermes cron (`hermes cron add --schedule "0 */12 * * *"
--script-only --command "hermes teams-pipeline maintain-subscriptions"`)
- Option 2: systemd service + timer (12h cadence, Persistent=true
so missed runs catch up after reboots)
- Option 3: plain crontab with a wrapper that sources .env for
credentials
Go-Live Checklist gains a bolded mandatory item for the schedule
being in place, with a cross-link to the section.
website/docs/user-guide/messaging/teams-meetings.md adds a
`::⚠️::` admonition right after the manual `subscribe`
examples so anyone who creates a subscription manually is told
the same day that it will silently expire in 72 hours.
2. Sidebar wiring. Shela's new docs pages (teams-meetings.md and
operate-teams-meeting-pipeline.md) weren't in website/sidebars.ts,
so they were orphaned URLs — reachable only if someone knew the
path. Wired teams-meetings into Messaging Platforms next to the
existing teams entry, and operate-teams-meeting-pipeline into
Guides & Tutorials next to microsoft-graph-app-registration from
PR #21922. Adjacent placement keeps the related pages discoverable
from each other.
3. SKILL.md rewrite (v1.0.0 → v1.1.0).
The original skill had five Turkish-only trigger phrases, which
works in a Turkish-speaking session but doesn't match English
triggers. Rewrote the skill to:
- Describe triggers by intent instead of exact phrases, with
explicit "works in any language" framing and example phrases
in both English and Turkish.
- Add a Decision Tree section covering the three most common user
asks (missing summary, setup verification, re-run request) and
the specific CLI command sequence for each.
- Add a dedicated "Critical pitfall: Graph subscriptions expire
in 72 hours" section that tells the agent exactly what to do
when a user reports "worked yesterday, nothing today" — the
most common operational failure mode.
- Expand the command reference into three labeled groups (Status
and inspection / Re-running and debugging / Subscription
management) so the agent can reach for the right command
without scanning.
- Add cross-links to all four related docs pages (Azure app
registration, webhook listener setup, full pipeline setup,
operator runbook).
Validation:
- npm run build: all new pages route, anchor to
#automating-subscription-renewal-required-for-production resolves
from both the runbook TOC and the teams-meetings.md admonition.
- scripts/run_tests.sh on the relevant test suites (607 tests): all
pass.
Background macOS desktop control via cua-driver MCP — does NOT steal the
user's cursor or keyboard focus, works with any tool-capable model.
Replaces the Anthropic-native `computer_20251124` approach from the
abandoned #4562 with a generic OpenAI function-calling schema plus SOM
(set-of-mark) captures so Claude, GPT, Gemini, and open models can all
drive the desktop via numbered element indices.
- `tools/computer_use/` package — swappable ComputerUseBackend ABC +
CuaDriverBackend (stdio MCP client to trycua/cua's cua-driver binary).
- Universal `computer_use` tool with one schema for all providers.
Actions: capture (som/vision/ax), click, double_click, right_click,
middle_click, drag, scroll, type, key, wait, list_apps, focus_app.
- Multimodal tool-result envelope (`_multimodal=True`, OpenAI-style
`content: [text, image_url]` parts) that flows through
handle_function_call into the tool message. Anthropic adapter converts
into native `tool_result` image blocks; OpenAI-compatible providers
get the parts list directly.
- Image eviction in convert_messages_to_anthropic: only the 3 most
recent screenshots carry real image data; older ones become text
placeholders to cap per-turn token cost.
- Context compressor image pruning: old multimodal tool results have
their image parts stripped instead of being skipped.
- Image-aware token estimation: each image counts as a flat 1500 tokens
instead of its base64 char length (~1MB would have registered as
~250K tokens before).
- COMPUTER_USE_GUIDANCE system-prompt block — injected when the toolset
is active.
- Session DB persistence strips base64 from multimodal tool messages.
- Trajectory saver normalises multimodal messages to text-only.
- `hermes tools` post-setup installs cua-driver via the upstream script
and prints permission-grant instructions.
- CLI approval callback wired so destructive computer_use actions go
through the same prompt_toolkit approval dialog as terminal commands.
- Hard safety guards at the tool level: blocked type patterns
(curl|bash, sudo rm -rf, fork bomb), blocked key combos (empty trash,
force delete, lock screen, log out).
- Skill `apple/macos-computer-use/SKILL.md` — universal (model-agnostic)
workflow guide.
- Docs: `user-guide/features/computer-use.md` plus reference catalog
entries.
44 new tests in tests/tools/test_computer_use.py covering schema
shape (universal, not Anthropic-native), dispatch routing, safety
guards, multimodal envelope, Anthropic adapter conversion, screenshot
eviction, context compressor pruning, image-aware token estimation,
run_agent helpers, and universality guarantees.
469/469 pass across tests/tools/test_computer_use.py + the affected
agent/ test suites.
- `model_tools.py` provider-gating: the tool is available to every
provider. Providers without multi-part tool message support will see
text-only tool results (graceful degradation via `text_summary`).
- Anthropic server-side `clear_tool_uses_20250919` — deferred;
client-side eviction + compressor pruning cover the same cost ceiling
without a beta header.
- macOS only. cua-driver uses private SkyLight SPIs
(SLEventPostToPid, SLPSPostEventRecordTo,
_AXObserverAddNotificationAndCheckRemote) that can break on any macOS
update. Pin with HERMES_CUA_DRIVER_VERSION.
- Requires Accessibility + Screen Recording permissions — the post-setup
prints the Settings path.
Supersedes PR #4562 (pyautogui/Quartz foreground backend, Anthropic-
native schema). Credit @0xbyt4 for the original #3816 groundwork whose
context/eviction/token design is preserved here in generic form.
Expand the google-workspace skill beyond read-only access to Drive and
Docs. Sheets already had full scope — just adds the missing create verb.
New subcommands:
- drive get : metadata for a single file
- drive upload : upload a local file (auto MIME detection)
- drive download : download or export (Docs/Sheets/Slides export to pdf/csv/pdf by default)
- drive create-folder
- drive share : user/group/domain/anyone + reader/writer/etc.
- drive delete : default trashes (reversible); --permanent skips the trash
- sheets create : new spreadsheet with optional first-tab name
- docs create : new doc, optional initial body
- docs append : append text at end of an existing doc
Scope changes:
- drive.readonly -> drive
- documents.readonly -> documents
Existing users with old tokens will hit the existing partial-scope
warning path (AUTHENTICATED (partial) ...) — the troubleshooting table
now points them at $GSETUP --revoke + redo steps 3-5 to pick up the
write scopes.
Small follow-ups on top of #19643:
- check_auth() takes quiet kwarg to suppress its AUTHENTICATED print
when called from check_auth_live(), so the final status line reflects
the live-call outcome only.
- Drop redundant _ensure_deps() call in check_auth_live() (check_auth()
already calls it).
- Add AUTHOR_MAP entry for ygd58 so release attribution script works.
setup.py --check only validated token shape/expiry but did not detect
when Google had disabled the OAuth client or account. Users got
AUTHENTICATED even when actual API calls failed with disabled_client.
Changes:
- Catch disabled_client and invalid_client in check_auth() refresh
path with actionable guidance (check Cloud Console, check account
status, do not retry)
- Add check_auth_live() that performs a real Calendar API call to
detect disabled_client errors that survive token refresh
- Add --check-live CLI flag backed by check_auth_live()
Fixes#19570
* feat(skills/linear): add Documents support + Python helper script
The bundled Linear skill (PR #1230) covered issues, projects, teams, and
workflow states via curl. It had no coverage for Linear's Documents API,
so fetching an RFC/doc from a linear.app URL required hand-writing
GraphQL against an underdocumented schema.
Adds:
- Documents section in SKILL.md explaining slugId extraction from URLs,
the contentState (markdown) vs contentState (ProseMirror) split, and
four canonical curl examples (fetch by slugId, fetch by UUID, list
recent, title-search).
- scripts/linear_api.py — stdlib-only Python CLI wrapping the most
common operations (whoami, list-teams, list/get/search/create/update
issues, add-comment, update-status, list/get/search documents, raw
GraphQL passthrough). Zero deps, reads LINEAR_API_KEY from env.
Auth header quirk (personal key takes bare $LINEAR_API_KEY, no Bearer
prefix) is already documented in the skill.
Found during RFC review: the existing skill's lack of document support
forced falling back to the browser (which hit Linear's login wall).
Also fixes a schema gotcha — the Document field is `contentState`, not
`contentData` (which returns 400).
Tested end-to-end against the production API:
python3 linear_api.py whoami
python3 linear_api.py get-document 38359beef67c
Both return expected payloads.
* fix(skills/linear): point LINEAR_API_KEY setup to the correct page
The org-level Settings > API page (/settings/api) only shows OAuth apps
and workspace-member keys. Personal API keys live under Account,
Security, access (/settings/account/security). Update both the setup
link in config.py (shown during hermes setup) and the setup step in
SKILL.md so users land on the page that can create a personal key.
Mirrors the AGENTS.md #20226 additions (Toolsets / Delegation / Curator /
Cron / Kanban) into the user-facing hermes-agent skill, and closes the
drift in the in-session slash command list.
User report (wxrrior in Discord): the skill did not mention /goal, so a
brand-new session answering "/hermes-agent do you have any info on /goal"
confidently said it did not exist. Cross-check against the CommandDef
registry found 16 commands missing from the static list: /goal, /agents,
/busy, /copy, /curator, /debug, /footer, /gquota, /indicator, /kanban,
/redraw, /reload, /reload-skills, /snapshot, /steer, /topic.
Changes:
- Slash Commands header now tells the reader to run /help or check the
live docs reference as the source of truth, and names the registry
of record (hermes_cli/commands.py) so future drift gets flagged
honestly instead of answered confidently wrong.
- Added all 16 missing commands, slotted into existing subsections
(/goal and /steer in Session; /busy + /indicator + /footer in
Configuration; /curator + /kanban + /reload-skills + /reload in
Tools & Skills; /topic in Gateway; /copy in Utility; /gquota +
/debug in Info).
- Toolsets table updated to the authoritative 30-key list from
toolsets.py (added kanban, yuanbao, spotify, safe, debugging, video,
feishu_doc, feishu_drive, discord, discord_admin, clarify; previously
stopped at 20 keys).
- New "Durable & Background Systems" section before Troubleshooting
covers Delegation, Cron, Curator, Kanban - each with a short rundown
of CLI verbs, key invariants, and a pointer to the user-facing docs.
Mirrors AGENTS.md #20226 but in the skill's user-facing register.
- Bumped version 2.0.0 -> 2.1.0.
Workers completing a kanban task can now claim the ids of cards they
created via an optional ``created_cards`` field on ``kanban_complete``.
The kernel verifies each id exists and was created by the completing
worker's profile; any phantom id blocks the completion with a
``HallucinatedCardsError`` and records a
``completion_blocked_hallucination`` event on the task so the rejected
attempt is auditable. Successful completions also get a non-blocking
prose-scan pass over their ``summary`` + ``result`` that emits a
``suspected_hallucinated_references`` event for any ``t_<hex>``
reference that doesn't resolve.
Closes#20017.
Recovery UX (kernel + CLI + dashboard)
--------------------------------------
A structural gate alone isn't enough — operators also need to see and
act on stuck workers, especially when a profile's model is the root
cause. This PR ships the full loop:
* ``kanban_db.reclaim_task(task_id)`` — operator-driven reclaim that
releases an active worker claim immediately (unlike
``release_stale_claims`` which only acts after claim_expires has
passed). Emits a ``reclaimed`` event with ``manual: True`` payload.
* ``kanban_db.reassign_task(task_id, profile, reclaim_first=...)`` —
switch a task to a different profile, optionally reclaiming a stuck
running worker in the same call.
* ``hermes kanban reclaim <id> [--reason ...]`` and
``hermes kanban reassign <id> <profile> [--reclaim] [--reason ...]``
CLI subcommands wired through to the same helpers.
* ``POST /api/plugins/kanban/tasks/{id}/reclaim`` and
``POST /api/plugins/kanban/tasks/{id}/reassign`` endpoints on the
dashboard plugin.
Dashboard surfacing
-------------------
* ⚠ **warning badge** on cards with active hallucination events.
* **attention strip** at the top of the board listing all flagged
tasks; dismissible per session.
* **events callout** in the task drawer — hallucination events render
with a red left border, amber icon, and phantom ids as styled chips.
* **recovery section** in the task drawer with three actions: Reclaim,
Reassign (with profile picker + reclaim-first checkbox), and a
copy-to-clipboard hint for ``hermes -p <profile> model`` since
profile config lives on disk and can't be edited from the browser.
Auto-opens when the task has warnings, collapsed otherwise.
Keyed by task id so state doesn't leak between drawers.
Active-vs-stale rule: warnings clear when a clean ``completed`` or
``edited`` event supersedes the hallucination, so recovery is never
permanently stigmatising — the audit events persist for debugging but
the badge goes away once the worker succeeds.
Skill updates
-------------
* ``skills/devops/kanban-worker/SKILL.md`` documents the
``created_cards`` contract with good/bad examples.
* ``skills/devops/kanban-orchestrator/SKILL.md`` gains a "Recovering
stuck workers" section with the three actions and when to use each.
Tests
-----
* Kernel gate: verified-cards manifest, phantom rejection + audit
event, cross-worker rejection, prose scan positive + negative.
* Recovery helpers: reclaim on running task, reclaim on non-running
returns False, reassign refuses running without reclaim_first,
reassign with reclaim_first succeeds on running.
* API endpoints: warnings field present on /board and /tasks/:id,
warnings cleared after clean completion, reclaim 200 + 409 paths,
reassign 200 + 409 + reclaim_first paths.
* CLI smoke: reclaim + reassign subcommands.
Live-verified end-to-end on a dashboard with seeded scenarios:
attention strip renders, badges land on the right cards, drawer
callout shows phantom chips, Reclaim on a running task flips status to
ready + emits manual reclaimed event + refreshes the drawer,
Reassign swaps the assignee and triggers board refresh.
359/359 kanban-suite tests pass
(test_kanban_{db,cli,boards,core_functionality} + dashboard + tools).
PR #9931 ("feat(google-workspace): add --from flag for custom sender display name")
accidentally removed the required_credential_files frontmatter block that tells
hermes to bind-mount google_token.json and google_client_secret.json into Docker
and Modal remote terminals before running setup.py.
Without this header the credential files are never registered in the session-scoped
ContextVar, so get_credential_file_mounts() returns an empty list at container
creation time and the OAuth files are invisible inside the sandbox.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The bundled himalaya skill documented folder aliases using a stale
TOML schema (`[accounts.NAME.folder.alias]`, singular) that himalaya
v1.2.0 silently ignores. The TOML parses without error, but the
alias resolver never reads the sub-section — every lookup then falls
through to the canonical folder name.
Source: in `pimalaya/core` (the `email-lib` crate himalaya v1.2.0
depends on, currently v0.27.0), `email/src/folder/config.rs` defines
`FolderConfig { aliases: Option<HashMap<String, String>>, ... }`
(plural, no `#[serde(rename)]`/`alias` aliases, no
`deny_unknown_fields`), and `account/config/mod.rs::get_folder_alias`
returns the input verbatim when no alias is found. So the singular
`alias` key deserializes to nothing and lookups silently fall
through.
On Gmail (where `sent` resolves to `[Gmail]/Sent Mail`, not `Sent`)
this means save-to-Sent fails *after* SMTP delivery already
succeeded, and `himalaya message send` exits non-zero. Any caller
(agent, script, user) that retries on that exit code will re-run
the entire send — including SMTP — producing duplicate emails to
recipients. Silent ignore + caller-level retry is significantly
worse than a config that just doesn't work.
This commit updates SKILL.md and references/configuration.md to the
v1.2.0 `folder.aliases.X` syntax (plural, dotted keys, directly
under the account section), adds a Gmail-specific block with the
`[Gmail]/Sent Mail`-style mapping, and adds notes on the failure
mode so future readers don't hit the same trap. SKILL.md version
bumped 1.0.0 → 1.1.0.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Moves the here-now skill under optional-skills/productivity/here-now/ so
it's discoverable via the Skills Hub but not installed by default, and
tightens the SKILL.md description to a single line to match sibling
optional-skill descriptions.
Install with:
hermes skills install official/productivity/here-now
Closes#378
Add the here.now productivity skill with a bundled publish runtime so Hermes can publish files and folders to live URLs. Keep the skill thin and docs-first while fixing script path resolution and upload failure handling.
Made-with: Cursor
Self-review caught several errors in the previous commit:
Frontmatter
- Replace non-standard `requires_runtime` / `requires_tooling` fields with
the documented `compatibility:` field (parsed by tools/skills_tool.py).
- Drop the `audit-v5` author tag I added unnecessarily.
MODEL_LOADERS catalog
- Remove `IPAdapterUnifiedLoader` (input `preset` is an enum, not a file).
- Remove `IPAdapterInsightFaceLoader` and `InsightFaceLoader` (input
`provider` is a GPU backend selector, not a model file). These would have
flagged enum values like "STANDARD" or "CUDA" as missing model files.
- Add "NB:" comment explaining `BasicGuider` has no `cfg` input
(the original PARAM_PATTERNS entry would never have matched).
- Remove `SamplerCustomAdvanced.noise_seed` from PARAM_PATTERNS — that
node takes a NOISE input from RandomNoise, not a seed field directly.
NODE_TO_PACKAGE registry slugs
- Verified all 18 packages against api.comfy.org and fixed:
- `comfyui-essentials` → `comfyui_essentials` (underscore, not hyphen)
- `comfyui-gguf` → `ComfyUI-GGUF` (case-sensitive)
- `comfyui-photomaker-plus` → `ComfyUI-PhotoMaker-Plus`
- `comfyui-wanvideowrapper` → `ComfyUI-WanVideoWrapper`
- ComfyUI-HunyuanVideoWrapper isn't on the registry; surface a git-URL
install hint via new NODE_TO_GIT_URL fallback so the user can install
via ComfyUI-Manager's /manager/queue/install endpoint.
Wrong class names
- `Canny` → `CannyEdgePreprocessor` (controlnet-aux registers the latter,
the former never appears in /object_info).
- Add `Zoe_DepthAnythingPreprocessor` and `AnimalPosePreprocessor` while
fixing controlnet-aux.
- Remove `Reroute (rgthree)` (rgthree's Reroute is JS-only — no Python
class, never appears in /object_info).
- Add `Display Int (rgthree)` (sibling of Display Any).
- Move `UltralyticsDetectorProvider` from `comfyui-impact-pack` to
`comfyui-impact-subpack` (separate package, registered there).
Tests
- Update test_packages_are_safe_for_shell to accept case-mixed slugs (the
registry uses both ComfyUI- and comfyui_ prefixes inconsistently). Replaced
the lowercase-only assertion with a shell-safe regex check.
- 117 tests still pass (105 unit + 8 cloud + 4 cross-host).
Attribution
- Add `SHL0MS@users.noreply.github.com` mapping to scripts/release.py
AUTHOR_MAP so check-attribution CI passes.
The audit of v4.1 surfaced ~70 issues across the five scripts and three
reference docs — most user-visible (silent file overwrites, status-error
misclassified as success, X-API-Key leaked to S3 on /api/view redirect,
Cloud endpoints that 404 because they were renamed). v5.0.0 fixes those
and fills the gaps that previously forced users to write their own glue
(WebSocket monitoring, batch/sweep, img2img upload helper, dep auto-fix,
log fetch, health check, example workflows).
Critical fixes
- run_workflow.py: poll_status now checks status_str==error BEFORE
completed:true, so a failed run no longer reports success
- run_workflow.py: download_output streams to disk via safe_path_join,
preserves server subfolder structure (no silent overwrites), and
retries with exponential backoff
- run_workflow.py: refuses to overwrite a link with a literal in
inject_params (would silently break wiring)
- _common.py: _StripSensitiveOnRedirectSession (subclasses
requests.Session.rebuild_auth) drops X-API-Key/Cookie on cross-host
redirects — fixes a real key-leak path through Cloud's signed-URL
download flow. Tested
- Cloud routing (verified live): /history → /history_v2,
/models/<f> → /experiment/models/<f>, plus folder aliases for the
unet ↔ diffusion_models and clip ↔ text_encoders rename
- check_deps.py: distinguishes 200/empty vs 404 folder_not_found vs
403 free-tier; emits concrete fix_command per missing dep
- extract_schema.py: prompt vs negative_prompt determined by tracing
KSampler.{positive,negative} connections (incl. through Reroute /
Primitive nodes) instead of meta-title heuristic; symmetric
duplicate-name resolution; cycle-safe trace_to_node
- hardware_check.py: multi-GPU pick-best, Apple variant detection,
Rosetta detection, WSL2, ROCm --json, disk-space check, optional
PyTorch probe; powershell preferred over deprecated wmic
- comfyui_setup.sh: prefers pipx → uvx → pip --user (with PEP-668
fallback); idempotent — skips relaunch if server already up;
configurable port/workspace; persistent log; SIGINT trap
New scripts
- run_batch.py — count or sweep (cartesian product), parallel up to
cloud tier limit
- ws_monitor.py — real-time WebSocket viewer; saves preview frames
- auto_fix_deps.py — runs comfy node install / model download for
whatever check_deps reports missing (with --dry-run)
- health_check.py — single command that runs the verification checklist
(comfy-cli + server + checkpoints + optional smoke test that cancels
itself to avoid burning compute)
- fetch_logs.py — pull traceback / status messages for a prompt_id
Coverage expansion
- Param patterns now cover Flux (BasicScheduler, BasicGuider,
RandomNoise, ModelSamplingFlux), SD3, Wan/Hunyuan/LTX video,
IPAdapter, rgthree, easy-use, AnimateDiff
- Embedding refs in CLIPTextEncode strings extracted as model deps
- ckpt_name / vae_name / lora_name / unet_name now controllable so
workflows can be retargeted per run
Examples
- workflows/{sd15,sdxl,flux_dev}_txt2img.json
- workflows/sdxl_{img2img,inpaint}.json
- workflows/upscale_4x.json
- workflows/{animatediff_video,wan_video_t2v}.json + README
Tests
- 117 tests (105 unit + 8 cloud integration + 4 cross-host security)
- Cloud tests auto-skip without COMFY_CLOUD_API_KEY; verified end-to-end
against live cloud API
Backwards compatibility
- All existing CLI flags continue to work; new behavior is opt-in
(--ws, --input-image, --randomize-seed, --flat-output, etc.)
Intended placement per PR #17610 discussion — comfyui belongs in
skills/creative/ alongside other creative built-ins (touchdesigner-mcp,
pretext, sketch), not in optional-skills/.
Pure directory rename, no content changes. History preserved via git mv.
Capture the reusable layout and animation lessons from the advanced Pretext demo so the skill teaches measured obstacle fields, morphing geometry, and polished browser examples.
Adds a 'pretext' skill under skills/creative/ for building cool browser
demos with @chenglou/pretext — the 15KB DOM-free text-layout library by
Cheng Lou.
The skill documents pretext as a creative primitive (not plumbing): text
flowing around obstacles, text-as-geometry games, proportional ASCII
surfaces, shatter/particle typography, editorial multi-column, kinetic
type, and multiline shrink-wrap. Each pattern pairs with copy-pasteable
snippets in references/patterns.md.
Two single-file HTML templates, both verified in a browser:
templates/hello-orb-flow.html
Minimal starter: long paragraph flows around a mouse-tracked orb
using layoutNextLineRange + a per-row corridor-width function.
templates/donut-orbit.html
Full 3D Sloane torus with orbit controls (drag to rotate, scroll to
zoom, idle auto-rotate). Each 'luminance pixel' is a real grapheme
sampled in reading order from a prose corpus via pretext's
prepareWithSegments + layoutWithLines + Intl.Segmenter. Amber-on-
black CRT aesthetic, z-buffer keyed by screen cell, 60fps.
Related skills: p5js, claude-design, excalidraw, architecture-diagram.
Background macOS desktop control via cua-driver MCP — does NOT steal the
user's cursor or keyboard focus, works with any tool-capable model.
Replaces the Anthropic-native `computer_20251124` approach from the
abandoned #4562 with a generic OpenAI function-calling schema plus SOM
(set-of-mark) captures so Claude, GPT, Gemini, and open models can all
drive the desktop via numbered element indices.
- `tools/computer_use/` package — swappable ComputerUseBackend ABC +
CuaDriverBackend (stdio MCP client to trycua/cua's cua-driver binary).
- Universal `computer_use` tool with one schema for all providers.
Actions: capture (som/vision/ax), click, double_click, right_click,
middle_click, drag, scroll, type, key, wait, list_apps, focus_app.
- Multimodal tool-result envelope (`_multimodal=True`, OpenAI-style
`content: [text, image_url]` parts) that flows through
handle_function_call into the tool message. Anthropic adapter converts
into native `tool_result` image blocks; OpenAI-compatible providers
get the parts list directly.
- Image eviction in convert_messages_to_anthropic: only the 3 most
recent screenshots carry real image data; older ones become text
placeholders to cap per-turn token cost.
- Context compressor image pruning: old multimodal tool results have
their image parts stripped instead of being skipped.
- Image-aware token estimation: each image counts as a flat 1500 tokens
instead of its base64 char length (~1MB would have registered as
~250K tokens before).
- COMPUTER_USE_GUIDANCE system-prompt block — injected when the toolset
is active.
- Session DB persistence strips base64 from multimodal tool messages.
- Trajectory saver normalises multimodal messages to text-only.
- `hermes tools` post-setup installs cua-driver via the upstream script
and prints permission-grant instructions.
- CLI approval callback wired so destructive computer_use actions go
through the same prompt_toolkit approval dialog as terminal commands.
- Hard safety guards at the tool level: blocked type patterns
(curl|bash, sudo rm -rf, fork bomb), blocked key combos (empty trash,
force delete, lock screen, log out).
- Skill `apple/macos-computer-use/SKILL.md` — universal (model-agnostic)
workflow guide.
- Docs: `user-guide/features/computer-use.md` plus reference catalog
entries.
44 new tests in tests/tools/test_computer_use.py covering schema
shape (universal, not Anthropic-native), dispatch routing, safety
guards, multimodal envelope, Anthropic adapter conversion, screenshot
eviction, context compressor pruning, image-aware token estimation,
run_agent helpers, and universality guarantees.
469/469 pass across tests/tools/test_computer_use.py + the affected
agent/ test suites.
- `model_tools.py` provider-gating: the tool is available to every
provider. Providers without multi-part tool message support will see
text-only tool results (graceful degradation via `text_summary`).
- Anthropic server-side `clear_tool_uses_20250919` — deferred;
client-side eviction + compressor pruning cover the same cost ceiling
without a beta header.
- macOS only. cua-driver uses private SkyLight SPIs
(SLEventPostToPid, SLPSPostEventRecordTo,
_AXObserverAddNotificationAndCheckRemote) that can break on any macOS
update. Pin with HERMES_CUA_DRIVER_VERSION.
- Requires Accessibility + Screen Recording permissions — the post-setup
prints the Settings path.
Supersedes PR #4562 (pyautogui/Quartz foreground backend, Anthropic-
native schema). Credit @0xbyt4 for the original #3816 groundwork whose
context/eviction/token design is preserved here in generic form.
Flips security.redact_secrets from true to false in DEFAULT_CONFIG, and
the HERMES_REDACT_SECRETS env-var fallback in agent/redact.py now
requires explicit opt-in ("1"/"true"/"yes"/"on") to enable.
New installs and users without a security.redact_secrets key get pass-
through tool output. Existing users whose config.yaml explicitly sets
redact_secrets: true keep redaction on — the config-yaml -> env-var
bridges in hermes_cli/main.py and gateway/run.py still honor their
setting.
Also updates the inline config comments, website docs, and the
hermes-agent skill so /hermes config set security.redact_secrets true
is now the documented way to turn it on.
Port https://github.com/blader/humanizer (MIT, v2.5.1, 16k stars) into
the built-in skills under skills/creative/humanizer/. Based on Wikipedia's
'Signs of AI writing' guide (WikiProject AI Cleanup) — detects 29 AI-writing
patterns and rewrites them to sound human.
Hermes-native adaptations:
- Description (<60 chars) explains what it's for: 'Humanize text: strip
AI-isms and add real voice.'
- 'When to use this skill' section — trigger phrases (humanize, de-AI,
de-slop, un-ChatGPT, rewrite to not sound like an LLM) plus guidance to
apply it to the agent's own output (release notes, PR descriptions, docs).
- 'How to use it in Hermes' — maps the three real input paths (inline,
file via read_file/patch/write_file, voice-calibration sample) onto the
tools the agent actually has. Drops Claude Code's allowed-tools block.
- Converted frontmatter to Hermes format (metadata.hermes.tags, category,
homepage, related_skills).
Attribution preserved:
- Original author Siqi Chen (@blader) credited in frontmatter and body.
- Full MIT LICENSE copied verbatim alongside SKILL.md.
- Wikipedia / WikiProject AI Cleanup credited.
- 29 patterns, personality/soul section, and full worked example kept
verbatim from the source (29,914 chars).
Validated end-to-end against a clean HERMES_HOME:
- sync_skills() copies skills/creative/humanizer/ including LICENSE.
- skills_list(category='creative') returns the 48-char description.
- skill_view(name='humanizer') returns the full body with all 29 patterns,
personality/soul, attribution, and Hermes tool refs (read_file, patch,
write_file) intact.
Adds four new reference docs covering common TD use cases not previously
documented in the skill:
- animation.md: LFOs, timers, keyframes, easing, time references
- midi-osc.md: MIDI controllers, OSC routing, TouchOSC, multi-machine sync
- particles.md: POPs and particleSOP — emission, forces, collisions, render
- projection-mapping.md: windowCOMP, corner pin, mesh warp, edge blending
Also clarifies the SKILL.md tool quick reference: adds td_screen_point_to_global
and notes that 4 admin/dev-mode tools (td_project_quit, td_test_session,
td_dev_log, td_clear_dev_log) live only in mcp-tools.md to keep the main
reference focused on creative workflows.
No SKILL.md workflow or critical-rules changes. References load on demand
so no token-budget impact at session start.
For 14 of 74 compressed skills, the original description contained
trigger keywords, technique counts, attribution, or use-case phrases
not covered by the existing body content. Prepends a 'When to use' /
'What's inside' block near the top so the agent still has the full
context when the skill is loaded.
Skills salvaged:
- codex, ascii-video, creative-ideation, excalidraw, manim-video, p5js
- gif-search, heartmula, youtube-content
- lm-evaluation-harness, obliteratus, vllm, axolotl
- powerpoint
Remaining 60 skills were verified to already cover the dropped content
in their existing body sections (When to Use, overview, intro prose)
or had short descriptions fully captured by the new compressed form.
Target: every skill's description fits in a one-line gateway menu and
leads with trigger keywords an agent would match on. Drops filler like
'Use this skill to', 'A skill for', 'This skill provides'.
Before: max description length was 791 chars (architecture-diagram),
74 of 81 built-in skills were >60 chars.
After: max 60, mean 54, all 81 built-in skills <=60.
Rewritten with double-quoted YAML scalars to preserve Chinese/arrow
glyphs (baoyu-comic, yuanbao, youtube-content).
- claude-design: 'Design one-off HTML artifacts (landing, deck, prototype).' (57)
- popular-web-designs: '54 real design systems (Stripe, Linear, Vercel) as HTML/CSS.' (60)
- design-md: "Author/validate/export Google's DESIGN.md token spec files." (59)
Also adds an inline callout near the top of claude-design pointing to
popular-web-designs and design-md so the cross-reference lands even
without reading the full decision table.
- claude-design: design process + taste for one-off HTML artifacts
- popular-web-designs: 54 ready-to-paste design systems (Stripe/Linear/etc.)
- design-md: formal DESIGN.md token spec file authoring
Adds a comparison table to claude-design's 'When To Use' section and
reciprocal pointers in design-md and popular-web-designs. Also corrects
claude-design author attribution to BadTechBandit.
Follow-up to #16323 — the UrlSource adapter is shipped but four
user-facing docs surfaces still only listed the hub-identifier forms.
- user-guide/features/skills.md: add ``url`` to the Supported-hub-sources
table; add a new "#### 8. Direct URL (`url`)" section explaining scope
(single-file SKILL.md only), name-resolution order (frontmatter → URL
slug → interactive prompt → --name flag), and both TTY and
non-interactive usage. Add two URL examples to the install-examples
block near the top of the page.
- reference/cli-commands.md: two URL install examples + one note
explaining the name-resolution fallback chain.
- guides/work-with-skills.md: one URL-install example alongside the
existing hub-identifier examples.
- skills/autonomous-ai-agents/hermes-agent/SKILL.md: Quick Reference
block's ``hermes skills install`` line now spells out that ID can be
a hub identifier OR a direct SKILL.md URL, and mentions --name for
frontmatter-less skills.
No code changes. No new dependencies. Website builds via the usual
Docusaurus pipeline.
Co-authored-by: teknium1 <teknium@noreply.github.com>
Parse scope from the raw callback URL before stripping the auth code so Flow.fetch_token matches user-granted scopes. Add regression test for dual-scope callbacks.
Made-with: Cursor
Four small tool-description / skill-content tweaks addressing recurring
model mistakes seen in @versun's docx feedback (Kimi 2.6, but the patterns
apply to every model):
1. browser_navigate description: call out .md/.txt/.json/.yaml/.csv/.xml,
raw.githubusercontent.com, and API endpoints as specifically preferring
curl or web_extract. The generic "prefer web_search or web_extract" was
too weak; models kept firing up the browser for plain-text URLs.
2. delegate_task description: two additions.
(a) Pass user language / output-style preferences in 'context' when they
differ from English — otherwise subagents default to English and their
summaries contaminate the final reply (caused the bilingual digest bug).
(b) Subagent summaries are self-reports, not verified facts. For
operations with external side-effects (HTTP uploads, remote writes,
file creation at shared paths), require a verifiable handle (URL, ID,
path) and verify it yourself before claiming success.
3. agent/prompt_builder.py Skills-mandatory block: new explicit line
"Whenever the user asks to configure / set up / modify / install /
enable / disable / troubleshoot Hermes Agent itself, load the
`hermes-agent` skill first." The generic "load what's relevant" didn't
route Hermes-meta questions (like "how do I turn off redaction?") to
the one skill that has the answer.
4. skills/autonomous-ai-agents/hermes-agent/SKILL.md: new "Security &
Privacy Toggles" section covering security.redact_secrets (with the
import-time-snapshot restart-required caveat), privacy.redact_pii,
approvals.mode (manual/smart/off) + --yolo + HERMES_YOLO_MODE, shell
hooks allowlist, and how to disable network/media tools entirely.
Every command verified against the actual config keys — no invented
knobs.
Co-authored-by: teknium1 <teknium@noreply.github.com>
Expand the airtable skill from bare CRUD to a full Hermes-shaped
cookbook matching the linear/notion neighbors, and trim the
description to fit the 60-char system-prompt cutoff.
Hermes-specific additions:
- Explicit 'use the terminal tool with curl — not web_extract or
browser_navigate' guidance, matching the same note in linear.
- Note that AIRTABLE_API_KEY flows from ~/.hermes/.env into the
subprocess automatically via env_passthrough, so curl calls don't
need to re-export it.
- Prefer 'python3 -m json.tool' (always present) over jq (optional)
for pretty-printing, with -s on every curl to keep output clean.
- Read-before-write workflow that resolves record IDs via
filterByFormula instead of guessing.
Cookbook expansion (new vs original):
- Field-type reference table (text, select, multi-select, attachment,
linked record, user) with the exact write-shape Airtable expects.
- typecast flag for auto-coercing values / auto-creating select options.
- performUpsert PATCH for idempotent sync by merge field.
- Batch create/delete endpoints (10-record cap per call).
- Sort + fields query params with URL-encoding (%5B / %5D).
- Named-view query that applies saved filter/sort server-side.
- Full pagination loop template (while loop with offset).
- Common filterByFormula patterns (exact match, contains, AND/OR,
date comparison, NOT empty).
- Rate-limit backoff guidance (Retry-After header, per-base budget).
- Airtable error-code reference (AUTHENTICATION_REQUIRED,
INVALID_PERMISSIONS, MODEL_ID_NOT_FOUND,
INVALID_MULTIPLE_CHOICE_OPTIONS) so the agent can map failures to
user-actionable fixes instead of just retrying.
Also: description trimmed from 183 chars (truncated to 60 in system
prompt, losing 'filter/upsert/delete' trigger terms) down to 59 chars
that render whole: 'Airtable REST API via curl. Records CRUD, filters,
upserts.' Catalog row updated to match.
SKILL.md grew from 115 to 228 lines — still under the 500-line soft
cap and below the linear skill (297 lines) which serves the same
role for GraphQL.
Convert the airtable skill from 'skills.config.airtable.api_key'
(config.yaml, wrong bucket for a secret) to 'prerequisites.env_vars:
[AIRTABLE_API_KEY]' (~/.hermes/.env), matching every other bundled
skill that authenticates with an API token.
Why the original shape was wrong:
- metadata.hermes.config is for non-secret skill settings (paths,
preferences) per references/skill-config-interface.md. Storing a
bearer token under skills.config.* also triggered the documented
'hermes config migrate' nag-on-every-run problem.
- The Quick Reference's 'AIRTABLE_API_KEY=...' bash line couldn't
read skills.config.airtable.api_key anyway — it's a yaml path, not
an env var.
Follow-up polish on the same pass:
- Added version/author/license frontmatter to match notion/linear.
- Added prerequisites.commands: [curl].
- Setup section now specifies the PAT format (pat...) that replaced
legacy 'key...' API keys in Feb 2024, plus the three required scopes
(data.records:read/write, schema.bases:read) and the per-base Access
list requirement.
- Clarified PATCH vs PUT and pagination (100 records/page cap).
- Swapped verification from 'hermes -q ...' (non-deterministic) to a
curl /v0/meta/bases call that returns a verifiable HTTP status code.
