The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting and analyzing external data (Pull Request metadata and source code), which creates a surface for indirect prompt injection if that data contains instructions designed to subvert the agent's behavior.
Ingestion points: Pull Request titles, bodies, and file paths are fetched via gh pr view in SKILL.md. File contents are accessed via the Read tool.
Boundary markers: The instructions do not define specific delimiters or "ignore embedded instructions" warnings when processing the external content.
Capability inventory: The skill utilizes Bash (to execute gh and grep), Read, and Glob tools.
Sanitization: There is no explicit sanitization or filtering of the fetched data before it is processed by the agent.
[COMMAND_EXECUTION]: The skill uses the Bash tool to interact with the GitHub CLI and search tools.
Evidence: The skill executes gh pr view to retrieve PR information and uses grep for pattern matching within the codebase as part of its architectural assessment protocol.

canon-design-review