The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources without adequate safeguards against malicious instructions.
Ingestion points: The skill fetches pull and merge request comments from GitHub, GitLab, and Azure DevOps (documented in SKILL.md and reference.md).
Boundary markers: There are no instructions to use delimiters or to warn the agent to ignore any instructions embedded within the fetched comments.
Capability inventory: The skill uses the gh, glab, and az command-line interfaces to interact with remote repositories.
Sanitization: There is no filtering or sanitization of the comment content before it is processed by the agent.
[COMMAND_EXECUTION]: The skill constructs shell commands using user-supplied parameters without explicit validation or sanitization instructions.
Instructions in SKILL.md and reference.md direct the agent to interpolate user-provided values, such as repository names, organization URLs, and PR/MR numbers, into shell commands (e.g., gh pr view <number> --repo org/repo).
While the agent is responsible for the execution, the lack of sanitization guidelines for these inputs represents a potential command injection surface.
[EXTERNAL_DOWNLOADS]: The skill references and suggests the installation of official command-line tools from well-known services.
It provides instructions for installing the GitHub CLI (gh), GitLab CLI (glab), and Azure CLI (az) via official domains and the Homebrew package manager (reference.md).
The skill follows best practices by instructing the agent to never install these tools without explicit user permission (SKILL.md).

aikido-mcp-comments