grimoire-improve
Pass
Audited by Gen Agent Trust Hub on Nov 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external source code, configuration files, and Git history, which constitutes an attack surface for indirect prompt injection.
- Ingestion points: The skill reads source code, Git history, configuration, and tests to perform its analysis as described in
SKILL.md. - Boundary markers: While explicit delimiters for the ingestion phase are not strictly defined in the instructions, the skill mandates robust output escaping.
- Capability inventory: The skill possesses capabilities to read files, execute Git history commands, and write report files.
- Sanitization: The instructions in
references/report-template.mdandreferences/mermaid-conventions.mdexplicitly direct the agent to escape all repository text and strictly avoid embedding secrets or executable content in its outputs. - [COMMAND_EXECUTION]: The skill utilizes Git history commands to identify hotspots and maintenance friction points.
- Evidence: The
SKILL.mdfile instructs the agent to "start discovery with recent Git change hotspots" and "Combine user-reported pain points with scoped Git history." - Context: This behavior is standard for the skill's primary purpose of architectural auditing and does not involve executing arbitrary or untrusted command strings.
Audit Metadata