The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses gh (GitHub CLI) and git commands to manage labels, search for issues, and create new feedback entries. These operations are restricted to the intended functionality of the skill.
[DYNAMIC_CONTEXT_INJECTION]: Employs the !command syntax to execute git remote -v and gh issue list at load time. This behavior is used to gather repository and issue context necessary for the skill's execution.
[DATA_EXFILTRATION]: While the skill sends data to GitHub by creating issues, it mitigates exfiltration risks by requiring the user to review and approve all findings via a multi-select prompt (AskUserQuestion) before any network requests are made.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted conversation history to generate feedback summaries and search queries.
Ingestion points: Step 2 analyzes the entire conversation history for feedback signals.
Boundary markers: None explicitly defined in the prompt instructions for the extraction phase.
Capability inventory: The skill can create GitHub issues and labels (gh issue create, gh label create) and perform git queries.
Sanitization: The skill relies on the agent's internal reasoning to filter findings, but does not provide specific sanitization instructions for the resulting issue titles or bodies.

feedback-session