The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources such as Linear task descriptions and GitHub PR content, which could contain malicious instructions designed to influence the agent's behavior during code modification or testing phases.
Ingestion points: The skill fetches issue details and comments using mcp__linear-server__get_issue and mcp__linear-server__list_comments in Phase 1. It also reads pull request data and diffs using gh pr view and gh pr diff in the Review mode.
Boundary markers: The instructions include a "Linear-fetching protocol" that verifies the issue title against a specific regex pattern ([multi-version][P##] ...) and checks the milestone. It also requires the agent to restate decisions and findings to the user.
Capability inventory: The skill uses the Bash tool to execute npx nx test and npx nx format, and the Write and Edit tools to modify source code in the repository.
Sanitization: The skill employs a mandatory "User OK gate" after Phase 2, where the agent must wait for explicit user approval before proceeding to Phase 3 (Implement), which involves branching and file edits.

multi-version-compliance