Commit graph

5 commits

Author SHA1 Message Date
Teknium
f2893fe51a
fix(tools): neutralize shell injection in _write_to_sandbox via path quoting (#7940)
_write_to_sandbox interpolated storage_dir and remote_path directly into
a shell command passed to env.execute(). Paths containing shell
metacharacters (spaces, semicolons, $(), backticks) could trigger
arbitrary command execution inside the sandbox.

Fix: wrap both paths with shlex.quote(). Clean paths (alphanumeric +
slashes/hyphens/dots) are left unmodified by shlex.quote, so existing
behavior is unchanged. Paths with unsafe characters get single-quoted.

Tests added for spaces, $(command) substitution, and semicolon injection.
2026-04-11 14:26:11 -07:00
adybag14-cyber
122925a6f2 fix(termux): honor temp dirs for local temp artifacts 2026-04-09 16:24:53 -07:00
alt-glitch
bbcff8dcd0 fix(tools): address PR review — remove _extract_raw_output, BudgetConfig everywhere, read_file hardening
- Remove _extract_raw_output: persist content verbatim (fixes size mismatch bug)
- Drop import aliases: import from budget_config directly, one canonical name
- BudgetConfig param on maybe_persist_tool_result and enforce_turn_budget
- read_file: limit=None signature, pre-read guard fires only when limit omitted (256KB)
- Unify binary extensions: file_operations.py imports from binary_extensions.py
- Exclude .pdf and .svg from binary set (text-based, agents may inspect)
- Remove redundant outer try/except in eval path (internal fallback handles it)
- Fix broken tests: update assertion strings for new persistence format
- Module-level constants: _PRE_READ_MAX_BYTES, _DEFAULT_READ_LIMIT
- Remove redundant pathlib import (Path already at module level)
- Update spec.md with IMPLEMENTED annotations and design decisions
2026-04-08 02:24:32 -07:00
alt-glitch
77c5bc9da9 feat(budget): make tool result persistence thresholds configurable
Add BudgetConfig dataclass to centralize and make overridable the
hardcoded constants (50K per-result, 200K per-turn, 2K preview) that
control when tool outputs get persisted to sandbox. Configurable at
the RL environment level via HermesAgentEnvConfig fields, threaded
through HermesAgentLoop to the storage layer.

Resolution: pinned (read_file=inf) > env config overrides > registry
per-tool > default. CLI override: --env.turn_budget_chars 80000
2026-04-08 02:24:32 -07:00
alt-glitch
65e24c942e wip: tool result fixes -- persistence 2026-04-08 02:24:32 -07:00