The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from GitHub issues (issueTitle, issueBody, and issueDetails). A maliciously crafted issue could attempt to influence the agent's behavior during the reproduction and fix implementation phases.
Ingestion points: issueTitle and issueBody are provided as arguments in SKILL.md; issueDetails is used in reproduce.md, diagnose.md, verify.md, and fix.md.
Boundary markers: No specific boundary markers or "ignore embedded instructions" warnings are used when processing the issue content.
Capability inventory: The skill executes shell commands (pnpm, node, git) and writes/executes scripts, as seen in reproduce.md and fix.md.
Sanitization: There is no explicit sanitization or validation of the content extracted from the GitHub issue.
[COMMAND_EXECUTION]: The skill uses the shell to perform several operations including installing dependencies (pnpm install), building the project (pnpm run build), running tests (pnpm test), and using the GitHub CLI (gh issue view).
[EXTERNAL_DOWNLOADS]: The reproduce.md file instructs the agent to run pnpm install, which fetches and installs Node.js dependencies from external registries.
[DYNAMIC_EXECUTION]: The reproduction process described in reproduce.md involves the agent programmatically creating a TypeScript reproduction script (test-repro.ts) and subsequently executing it using node.

triage