The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data originating from GitHub Pull Requests and Issues, creating a surface for indirect prompt injection attacks.
Ingestion points: The skill ingests external content in SKILL.md (Step 1 and Step 4) using gh pr view, gh issue view, and git diff to retrieve metadata, comments, and code changes.
Boundary markers: The instructions do not define explicit boundary markers or delimiters to separate the untrusted data from the agent's instructions, nor do they include warnings to ignore instructions embedded in the analyzed content.
Capability inventory: The skill has access to the GitHub CLI (gh) for repository operations and git for version control, including the ability to clone repositories and check out code.
Sanitization: There is no evidence of sanitization, escaping, or filtering of the external content before it is processed by the agent.
[COMMAND_EXECUTION]: The skill relies on several shell commands to interact with the environment and perform the code review.
Evidence: The instructions in SKILL.md utilize gh for GitHub metadata and repository management, git for calculating diffs and checking out branches, and mktemp to create temporary directories for cloning target repositories.
[EXTERNAL_DOWNLOADS]: The skill fetches external content from GitHub as part of its core functionality.
Evidence: The skill uses gh repo clone and gh pr checkout to download code from specified GitHub repositories for review. These operations target a well-known service (GitHub) and are consistent with the skill's stated purpose of reviewing pull requests.

nemoclaw-maintainer-security-code-review