The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches a third-party package from the NPM registry at runtime during the review subscription step.
Evidence: npx -y crysk status --pr <n> --wait --json in SKILL.md.
[REMOTE_CODE_EXECUTION]: The skill executes the code downloaded via npx immediately after fetching, which occurs without explicit version pinning or integrity verification.
Evidence: npx -y crysk status command execution in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from the repository and external tool output to determine its next autonomous steps in the PR workflow.
Ingestion points: Reads repository-level PR templates (.github/pull_request_template.md) and configuration files (CONTRIBUTING.md, AGENTS.md). It also captures the JSON output from the crysk status command.
Capability inventory: git push, gh pr create, npx -y crysk, and the ability to load and execute the follow-up crysk-loop skill.
Boundary markers: The instructions do not define specific delimiters or warnings to ignore instructions that might be embedded within the ingested templates or tool-generated JSON reports.
Sanitization: No sanitization or validation logic is present for handling these external inputs before they influence the agent's decision-making loop.
[DYNAMIC_EXECUTION]: The skill dynamically transitions to and executes another skill based on the state reported by an external tool.
Evidence: Immediately load the crysk-loop skill (skills/crysk-loop/SKILL.md) and run its autonomous loop instructions in SKILL.md.
[COMMAND_EXECUTION]: The skill executes several shell commands to manage branches, push code, and create pull requests.
Evidence: Commands include git push, gh repo view, gh pr create, and gh pr view.

crysk-pr