The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon data from external GitHub issues, which represents a potential indirect prompt injection surface.
Ingestion points: Fetches issue titles, bodies, and comments using the gh-fetch-issue utility (referenced in SKILL.md Step 1b and references/issue-sweep.md Step 2).
Boundary markers: Relies on the ⚠️ UNTRUSTED fencing mechanism provided by gh-fetch-issue to distinguish between trusted collaborator content and untrusted third-party input.
Capability inventory: The skill possesses extensive capabilities, including creating/editing GitHub issues, managing Git branches/remotes, and writing logs to the local filesystem.
Sanitization: Explicitly instructs the agent to treat fenced content as data only, strictly prohibiting the derivation of plan steps or execution of commands from untrusted blocks.
[COMMAND_EXECUTION]: The skill performs numerous command-line operations to facilitate its workflow.
Executes git for branch detection, creation, and remote synchronization.
Uses gh CLI for comprehensive GitHub issue and label management.
Invokes node and python3 to run internal utility scripts for logging, validation, and orientation tracking.

big-plan