The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to perform reconnaissance on the local environment by executing directory listings (ls) on internal agent paths (e.g., ~/.claude/skills/, ~/.agent/skills/) to identify other software components. It further directs the agent to suggest installation of additional components using npx.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface because it processes external data. 1. Ingestion points: Data is loaded from data.csv using pd.read_csv() in SKILL.md. 2. Boundary markers: Absent; the skill does not use delimiters to isolate data from instructions. 3. Capability inventory: Includes filesystem listing (ls) and Python code execution. 4. Sanitization: Absent; no input validation is performed on the ingested data.

data-science