ac-commit

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes Git diffs, which are considered untrusted external data, to generate commit messages and context entries.
  • Ingestion points: Untrusted content is ingested via git diff --cached and git diff during the analysis and archiving phases.
  • Boundary markers: The instructions do not define specific delimiters or boundary markers to isolate the diff content from the sub-agent instructions.
  • Capability inventory: The skill is authorized to perform Git commits, staging, and local file writes to the .context/ directory within the repository.
  • Sanitization: The skill includes a proactive security step in Phase 5 to scan for sensitive patterns (tokens, keys, and passwords) and redact them using a [REDACTED] placeholder before writing to the history logs.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 09:46 AM