The Agent Skills Directory

[PROMPT_INJECTION]: No patterns detected that attempt to override system safety or ignore instructions. The instructions manage the session termination process without requesting safety bypasses.- [DATA_EXFILTRATION]: No network calls or sensitive file access observed. The skill uses git status and git diff for reporting purposes only and does not exfiltrate data to external domains.- [REMOTE_CODE_EXECUTION]: No remote downloads or code execution patterns detected. Operations are limited to local version control and agent messaging.- [OBFUSCATION]: No encoded strings, homoglyphs, or hidden text identified in the instructions or configuration.- [INDIRECT_PROMPT_INJECTION]: The skill processes external inputs to synthesize a handoff. Ingestion points: user-provided arguments, worker subagent reports, and git tree inventory. Capability inventory: git commit authorization and inter-agent communication tools. Boundary markers: The skill enforces a specific structured format for worker reports. Sanitization: No explicit content filtering is defined, but the data is constrained to administrative reporting.

wrap-up