The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several system commands to gather and manage session state, including git status, git diff, git log, and mktemp. These are routine operations for a session management tool and use secure practices like avoiding hardcoded temporary paths.
[EXTERNAL_DOWNLOADS]: The skill interacts with the GitHub API via the gh CLI to retrieve issue content in the resume mode. This is used to reconstruct the context of a previous work session. Interaction is limited to the repository context.
[DATA_EXFILTRATION]: The create mode gathers repository information and conversation history to create or update GitHub issues. To protect sensitive data, the skill instructions and templates explicitly mandate the redaction of secrets, tokens, and PII before any data is sent to GitHub.
[PROMPT_INJECTION]: The resume mode presents an indirect prompt injection surface as it ingests data from external GitHub issues which could theoretically be manipulated.
Ingestion points: GitHub issue bodies are read via gh issue view and gh issue list in SKILL.md.
Boundary markers: The skill validates content using a kind: handoff YAML metadata block, providing a basic organizational boundary.
Capability inventory: The skill has access to shell execution (git, gh, claude), local file system access, and the ability to spawn background agents (claude --bg).
Sanitization: No rigorous sanitization or filtering is applied to the retrieved issue body content prior to processing, relying instead on the metadata check and user verification of state drift.

handoff