The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the gh CLI and git commands to interact with GitHub repositories. The Python script scripts/create_issues.py uses subprocess.run to execute these system binaries for issue management.
[CREDENTIALS_UNSAFE]: The skill includes instructions and code to extract GitHub access tokens from the local git configuration (git remote get-url). While used for legitimate automation within the author's workflow, this pattern programmatically accesses sensitive authentication data.
[DATA_EXFILTRATION]: The skill transmits session-derived data (issue titles and bodies) to an external GitHub repository (minicoohei/ai-agent-camp). This aligns with the skill's stated purpose and targets a well-known service associated with the author.
[PROMPT_INJECTION]: The skill processes session history to generate issues, which presents a surface for indirect prompt injection. However, the workflow includes a mandatory user confirmation step ('Phase 3: User Confirmation') where the user must approve the content before any issues are registered, significantly mitigating the risk of unauthorized or malicious actions.

session-retrospective