The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a bash script to filter and load session files: bash ${CLAUDE_PLUGIN_ROOT}/scripts/retrospect-load-sessions.sh $@. This is a standard procedure for internal plugin logic. The script location is scoped to a plugin root variable, which is a safe practice.
[REMOTE_CODE_EXECUTION]: No remote code execution or untrusted downloads were found. The skill relies on local session files and internal scripts.
[DATA_EXFILTRATION]: The skill writes analysis results to a local, project-configurable directory (.retro/insights/). It does not contain any network operations or external data exfiltration patterns.
[CREDENTIALS_UNSAFE]: No hardcoded secrets, API keys, or sensitive credential file paths were found.
[INDIRECT_PROMPT_INJECTION]: The skill processes session data, which is an ingestion surface. However, it includes a clear "Extract, don't invent" principle and instructions to only base insights on evidenced session content, reducing the risk of accidental obedience to instructions embedded in the analyzed data.

retrospect-domain