The Agent Skills Directory

[NO_CODE]: The skill relies primarily on a static reference library (66 Markdown files) and a simple Python script for local file searching.
[REMOTE_CODE_EXECUTION]: No remote dependencies or external script execution patterns were found. The included script scripts/lookup.py uses only Python standard library modules and does not perform network operations.
[DATA_EXFILTRATION]: No evidence of network activity, credential harvesting, or unauthorized file access was detected. The workflow strictly limits the agent's behavior to local reference lookups.
[PROMPT_INJECTION]: The instructions in SKILL.md are designed to reinforce safety and clinical boundaries. There are no patterns suggesting attempts to bypass AI safety guardrails; instead, they include explicit 'Non-negotiable rules' prioritizing crisis intervention and safety.
[DYNAMIC_CONTEXT_INJECTION]: The skill does not utilize dynamic shell command execution placeholders in its SKILL.md frontmatter.
[INDIRECT_PROMPT_INJECTION]: While the skill processes user-supplied symptom descriptions, it mitigates risk through a strictly defined workflow that requires the agent to utilize specific local references and maintain a 'candidate condition' framing with uncertainty markers. The attack surface is minimal as the skill lacks high-privilege capabilities like network access or file system writes.

dsm5