The Agent Skills Directory

[SAFE]: The skill does not include any executable scripts, binaries, or third-party dependencies, relying entirely on markdown instructions for its functionality.
[COMMAND_EXECUTION]: The skill utilizes standard read-only commands such as rg, git diff, git log, and git show to inspect the repository. These tools are used appropriately for debugging and investigation tasks within the user's workspace.
[PROMPT_INJECTION]: The skill processes untrusted data from logs, crash traces, and git history, which represents an indirect prompt injection surface. This is a low-risk finding as the skill enforces strict read-only constraints and prohibits file modification. 1. Ingestion points: Application logs, crash traces, and repository history (SKILL.md). 2. Boundary markers: The skill does not use specific delimiters for external content but explicitly enforces 'read-only' instructions for all sub-agents. 3. Capability inventory: File reading and git inspection tools; no file-writing or network-transmission capabilities are requested. 4. Sanitization: No explicit sanitization or validation of the ingested content is performed.

bug-hunt-swarm