The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: In Step 4 of SKILL.md, the skill executes project test suites (such as rspec, pytest, go test, or npm test) and repository hooks on code checked out from remote pull request branches using the gh pr checkout command. This results in the execution of unverified code from remote sources in the user's local environment.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection from pull request metadata. Ingestion points: Pull request titles, bodies, and review comments are fetched in SKILL.md Step 2. Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present. Capability inventory: The skill has access to the Bash tool, file system operations (Read, Write, Edit), and the ability to invoke external skills like wk-workflow. Sanitization: No sanitization or validation of the fetched metadata is performed. Malicious instructions placed in PR comments or descriptions could influence the agent's orientation and the task list generated in Step 6.

wk-pr-takeover