The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external tracking systems and public repositories, which may contain untrusted content or instructions designed to influence the agent's behavior.
Ingestion points: The skill fetches context using gh issue view, gh pr view, gh api (for repository and public security advisories), and a Linear integration. It also processes external video transcripts.
Boundary markers: The skill wraps user-provided task arguments in a <task> XML tag to differentiate input from instructions.
Capability inventory: The agent has the capability to read/write files, execute shell commands (git, gh, npm, pnpm, bun, node), and load specialized supplementary skills.
Sanitization: The skill mitigates risks by requiring a 'Public Issue Challenge Gate', which mandates that the agent adversarialy review external diagnoses and reproduce reported bugs before implementing changes.
[COMMAND_EXECUTION]: The skill utilizes several command-line utilities to perform development workflows, which are powerful capabilities that could be targeted by an attacker.
Evidence: Instructions include the use of git for branch management, gh for interacting with GitHub APIs, npm/pnpm/bun for package management and testing, and executing repository-local scripts like create-goal-scratchpad.mjs.
Context: These tools are used for their intended purposes within a software development environment and are documented as part of the core execution path for tasks like bug fixing, testing, and documentation.

task