The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses a significant surface area for indirect prompt injection. Its core functionality involves reading and analyzing untrusted external documents, such as ToRs (Terms of Reference), RFPs (Requests for Proposals), and adverts. Because the agent is granted the capability to write files and execute local scripts, a maliciously crafted document could potentially override the agent's instructions or trigger unauthorized actions.
Ingestion points: Main SKILL.md and ai-agent-discovery-and-qualification/SKILL.md direct the agent to read external files (ToR, RFP, adverts).
Boundary markers: The instructions do not define explicit delimiters or use 'ignore embedded instructions' warnings for the external content.
Capability inventory: The agent has permission to read, search, and edit markdown files, and to execute several bundled Python scripts.
Sanitization: The skill relies on 'slop' audits which focus on writing style rather than security sanitization or filtering of external inputs.
[COMMAND_EXECUTION]: The skill includes a set of local Python utility scripts (scripts/init_skill.py, scripts/package_skill.py, and scripts/quick_validate.py) and instructs the agent to execute them for environment initialization and validation. While these specific scripts appear to perform benign management tasks, the execution of local code alongside the ingestion of untrusted data constitutes an increased security risk.

skills