The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon GitHub review comments, which are external, untrusted inputs that could be controlled by a malicious actor. This creates a potential surface for instructions embedded in review comments to influence the agent's behavior.
Ingestion points: The scripts/get-review-comments.sh script and assets/pull-request-review-threads.graphql are used to fetch comment bodies from GitHub into the agent's context.
Boundary markers: The skill instructs the agent to "understand, research, and judge each comment" and classify it, but it lacks strict technical delimiters to isolate the comment content from the agent's operational instructions.
Capability inventory: The skill allows the agent to modify source code in the repository, execute validation commands, create git commits, and push changes to the remote repository.
Sanitization: There is no documented logic for sanitizing or escaping the content of the review comments before they are processed by the agent.
[COMMAND_EXECUTION]: The skill relies on shell scripts and command-line tools to perform its operations, which involves executing commands on the local system.
Evidence: The skill executes scripts/get-review-comments.sh, scripts/resolve-review-comments.sh, git, and gh CLI tools.
Safety Measures: The provided bash scripts include regex-based validation for owner and repo arguments to mitigate shell command injection risks.

fusion-github-review-resolution