The Agent Skills Directory

[SAFE]: The skill functions as a meta-cognitive framework for maintaining session continuity and identity coherence. It does not exhibit malicious behaviors or attempt to bypass security guidelines.
[COMMAND_EXECUTION]: The skill instructions include executing git log and ls -lt to identify recent project activity. These are read-only commands used for context reconstruction and do not pose a privilege escalation or system integrity risk.
[PROMPT_INJECTION]: An indirect prompt injection surface is present (Category 8). 1. Ingestion points: Git commit messages (Step 2.1) and source code comments found via Grep (Step 2.3). 2. Boundary markers: Absent; the instructions integrate findings directly into the agent's reconstructed context. 3. Capability inventory: Read, Grep, and Glob tools are authorized. 4. Sanitization: Absent; the agent integrates evidence from artifacts without explicit filtering. This is a low-risk architectural feature necessary for the skill's primary function.

bootstrap-agent-identity