janitor-discover
Pass
Audited by Gen Agent Trust Hub on Aug 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local bash scripts (
discover.sh,search.sh,precheck.sh,security.sh) located in the plugin's script directory to perform its primary management functions. - [EXTERNAL_DOWNLOADS]: Fetches manifest files (SKILL.md) from external GitHub repositories via the GitHub API and direct URLs to facilitate pre-installation analysis.
- [PROMPT_INJECTION]: The skill processes untrusted external data, which creates a surface for indirect prompt injection.
- Ingestion points: Fetches and reads
SKILL.mdcontent from arbitrary GitHub repositories provided by the user or found via search. - Boundary markers: Analysis results are presented in delimited blocks (e.g.,
--- Security (<scope>) ---) to separate external content from system instructions. - Capability inventory: The environment allows execution of local bash scripts and network operations via
curl. - Sanitization: The skill uses a dedicated heuristic scanner (
security.sh) to detect and report suspicious patterns in external manifests before they are presented to the agent or user.
Audit Metadata