running-retro

Pass

Audited by Gen Agent Trust Hub on Jul 30, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill implements a retrospective analysis of session transcripts, which are untrusted data sources (Ingestion points: SKILL.md Step 2). Potential risks are mitigated by a 'Trust boundary' warning in the delegation prompt in context/checkpoint.md (Boundary markers). Capabilities are limited to reading local files and running a local Python parser (Capability inventory). Mandatory redaction sweeps are performed on all outputs before they are appended to the ledger (Sanitization).
  • [DYNAMIC_CONTEXT_INJECTION]: The SKILL.md file uses dynamic shell execution to collect environment metadata such as git branch status and commit history. These operations are strictly read-only, do not incorporate user-controlled input, and provide necessary context for the retrospective task without introducing security risks.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 30, 2026, 07:55 PM
Security Audit — agent-trust-hub — running-retro