retro

Pass

Audited by Gen Agent Trust Hub on Apr 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes Bash shell commands to calculate metrics, including git, grep, and wc. It also executes dynamically determined 'project-specific' build and test commands, which could be exploited to run arbitrary code if the repository being analyzed contains malicious configurations.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it explicitly directs the agent to locate a 'LESSONS.md' file and 'apply those project-specific lessons' alongside its own internal rules.
  • Ingestion points: Project artifacts including LESSONS.md, INTAKE.md, PLAN.md, BUILD-LOG.md, and REVIEW.md.
  • Boundary markers: There are no delimiters or protective instructions used to separate the content of these external files from the agent's core instructions.
  • Capability inventory: The skill is granted extensive capabilities including shell access (Bash) and file manipulation tools (Read, Write, Edit, Glob, Grep).
  • Sanitization: Content read from project artifacts is used directly without sanitization or validation to influence the retrospective report and the agent's subsequent reasoning.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 7, 2026, 02:59 AM
Security Audit — agent-trust-hub — retro