The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data from session logs, which creates an indirect prompt injection surface. Content within the logs could potentially influence the audit report or the removal suggestions provided to the agent.
Ingestion points: Session logs (.jsonl files) in ~/.pi/agent/sessions/ parsed by scripts/audit.py.
Boundary markers: Absent; the instructions do not provide delimiters or warnings to ignore instructions embedded within the scanned log data.
Capability inventory: The agent has the capability to execute shell commands (bunx skills remove) and read local file contents.
Sanitization: The script extracts skill names via regular expressions but does not perform validation or escaping of the strings before they are presented in removal command templates.
[COMMAND_EXECUTION]: The skill is designed to guide the agent in executing administrative commands (bunx skills remove) to manage the local environment. This behavior is consistent with the skill's primary purpose of skill management and requires user confirmation.
[SAFE]: The skill performs its functions locally without making network requests or exfiltrating data. It does not contain obfuscated code, persistence mechanisms, or unauthorized privilege escalation attempts.

pi-skill-audit