feature-docs

Pass

Audited by Gen Agent Trust Hub on Oct 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONPERSISTENCEINDIRECT_PROMPT_INJECTIONPRIVILEGE_ESCALATION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses a local bash script (scripts/feature-docs.sh) to perform file management, hashing (using shasum or sha256sum), and git repository interaction. It utilizes standard Unix utilities like sed, awk, find, and diff to process documentation files and generate summaries.
  • [PERSISTENCE]: The skill includes a command (hook) to install a git pre-commit hook in the local repository's .git/hooks/ directory. It also recommends adding a SessionStart hook to the agent's configuration file (.claude/settings.json) to automate document status checks. These are intended persistence mechanisms for developer workflow automation.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from untrusted or local sources, such as .scratch/ directories and existing project documentation. This represents an attack surface where maliciously crafted local content could attempt to influence the agent's documentation output or script logic.
  • Ingestion points: .scratch/<feature>/ (spec and issue files) and docs/<feature>/feature-docs/ (existing documentation).
  • Boundary markers: Absent in prompt interpolation logic.
  • Capability inventory: Shell script execution, file read/write, git hook installation, and permission modification (chmod +x).
  • Sanitization: The script implements basic character filtering for filenames and paths to prevent simple command injection, but relies on the user providing valid paths.
  • [PRIVILEGE_ESCALATION]: During the installation of git hooks, the script executes chmod +x on the newly created pre-commit file. This modification is restricted to the local development environment and is a standard requirement for the hook to function as intended.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 7, 2026, 02:06 PM