The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the gh-stack extension from the official github organization repository using gh extension install github/gh-stack. This is a download from a well-known service.\n- [COMMAND_EXECUTION]: The skill utilizes several Git and GitHub CLI commands to manipulate repository state, including git config, git commit, git rebase, and various gh stack subcommands. These are necessary for the skill's functionality.\n- [INDIRECT_PROMPT_INJECTION]: The skill exposes the agent to untrusted data that could contain malicious instructions.\n
Ingestion points: The agent reads PR metadata (titles, bodies, branch names) via gh stack view --json and commit history via git log.\n
Boundary markers: No delimiters or defensive instructions are provided to the agent to distinguish between data and instructions in these inputs.\n
Capability inventory: The agent can execute shell commands, perform network operations (via gh stack push/submit), and modify repository configuration.\n
Sanitization: The skill lacks sanitization routines for the ingested data before it is used in command arguments or decision logic.\n- [METADATA_POISONING]: The SKILL.md file identifies the author as github, whereas the skill is provided by another author. While this could be misleading, it likely reflects the authorship of the gh-stack tool itself.

gh-stack