The Agent Skills Directory

[PROMPT_INJECTION]: The skill employs authoritative directives such as "CRITICAL," "MANDATORY," "Rule," and "VIOLATION" to override the agent's default behavioral patterns. It specifically commands the agent to be "concise, direct, and solution-focused," and instructs the agent to fix bugs without explanation, which overrides standard conversational transparency.
[COMMAND_EXECUTION]: The skill mandates the execution of various Python scripts located in the local environment directory ~/.claude/skills/ (e.g., python ~/.claude/skills/vulnerability-scanner/scripts/security_scan.py .). It provides a mapping of specific agents to these scripts and requires their execution to verify tasks before completion.
[INDIRECT_PROMPT_INJECTION]: The skill's workflow involving the processing of project source code and tool outputs presents an attack surface for indirect prompt injection.
Ingestion points: The agent is instructed to read and edit project files and to "Run the script and capture ALL output" from the verification tools.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands in the processed files or tool outputs are provided.
Capability inventory: The skill uses Read, Write, and Edit tools and mandates the execution of local shell commands via Python scripts.
Sanitization: No sanitization or validation logic is specified for the external data ingested during the audit and editing process.

clean-code