The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which could contain malicious instructions designed to manipulate the agent.
Ingestion points: The skill fetches PR diffs, reviewer comments, and CI logs in Step 1 using FetchUrl or gh pr view.
Boundary markers: The instructions do not define delimiters or specific "ignore embedded instructions" warnings for the fetched data.
Capability inventory: The skill has capabilities to execute shell commands via git, gh, and various language-specific build/test tools (Step 5 and Step 7).
Sanitization: There is no mention of sanitizing or escaping the content fetched from the PR before the agent processes it to "address reviewer comments".
[COMMAND_EXECUTION]: The workflow relies on executing multiple CLI tools based on external input and repository state.
The skill uses git for branch management and gh (GitHub CLI) for PR interactions and metadata retrieval.
Step 5 encourages the execution of project-specific build, lint, and test commands which are variable and dependent on the repository's configuration.
[DYNAMIC_EXECUTION]: The skill dynamically determines which commands to execute by reading project configuration files at runtime.
In Step 5, the agent is instructed to discover build/test commands by reading files like package.json, pyproject.toml, Makefile, etc., and then execute the discovered commands (e.g., npm run test, pytest, cargo test). This is a standard part of the skill's developer-focused functionality but represents a dynamic execution pattern.

follow-up-on-pr