The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from local specification files and GitHub issues, which could theoretically contain malicious instructions designed to influence the agent's behavior.
Ingestion points: The skill reads local files using Read, Grep, and Glob tools, and fetches remote GitHub issue data using the read_issue operation as described in SKILL.md.
Boundary markers: The skill mandates a "Mandatory Draft Gate" where the agent must present all proposed changes to the user for explicit confirmation before any mutations occur.
Capability inventory: The skill uses Bash for shell operations, temporary file management, and gh CLI execution, alongside the mcp__github tool for GitHub API interactions.
Sanitization: The instructions explicitly require the use of quoted shell heredocs (<<'EOF') in the references/github-surface.md file when generating temporary markdown files, which prevents the shell from expanding variables or executing commands embedded within the task content.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform environment capability checks, manage temporary files for issue bodies, and execute the gh CLI for GitHub interactions when the tool is available.

plan-to-graph