skill-review
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides instructions and shell scripts for auditing the file system. These include using
findto identify broken symlinks,ln -sfor creating canonical links, and a project-specific CLI tool (joelclaw) to send events and ensure skill integrity. These operations are limited to specific local directories related to the agent's configuration. - [INDIRECT_PROMPT_INJECTION]: The skill describes an automated process ('Skill Garden') that ingests system state and local files like
AGENTS.mdto perform content audits. - Ingestion points: Reads
AGENTS.mdand checks file system patterns for stale architecture references. - Boundary markers: None explicitly defined in the provided markdown, though it functions as a diagnostic tool.
- Capability inventory: Involves file system reads (
find,read), repository management (git), and symlink creation (ln -s). - Sanitization: None detected in the documentation; however, the skill is intended for trusted administrative use within a managed repository.
Audit Metadata