The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the GitHub Command Line Interface (gh) to manage issues and projects.
Evidence: Multiple shell commands including gh issue edit, gh issue comment, and gh project item-edit are documented for use in Step 10 and the GitHub Adapter section.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection, an inherent risk for agents processing external tracker data.
Ingestion points: Data enters the context via GitHub Issues, GitHub Projects, Linear boards, and local markdown files (README.md, CLAUDE.md, backlog files) as described in Steps 1, 2, and 4.
Boundary markers: None identified. There are no instructions to wrap ingested issue content in delimiters or ignore embedded instructions within issue bodies.
Capability inventory: The agent has the authority to edit labels, post comments, close issues, and modify project board statuses.
Sanitization: None identified. The skill does not describe any validation or filtering logic for the text retrieved from issue trackers.
[SAFE]: The skill includes several security-positive constraints.
It defaults all operations to dry-run to prevent accidental mutations.
It explicitly forbids the deletion of branches, limiting itself to reporting cleanup candidates.

backlog-manager