The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a suite of local helper binaries to manage its workflow and gather metrics.
Evidence: Calls to gstack-skill-start, gstack-retro-metrics, gstack-global-discover, gstack-learnings-search, and gstack-question-log located in ~/.claude/skills/gstack/bin/.
[COMMAND_EXECUTION]: The skill interacts with external hosting platforms to fetch repository metadata.
Evidence: Uses git fetch, gh pr list, and glab mr view to query GitHub and GitLab for merged pull requests and commit history.
[DYNAMIC_EXECUTION]: The skill uses dynamic execution patterns to configure its environment and locate components.
Evidence: Uses eval to source the output of the local gstack-slug binary into the shell environment.
Evidence: Dynamically locates and executes a discovery script using bun run bin/gstack-global-discover.ts if a compiled binary is not present.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it processes data from untrusted sources.
Ingestion points: Git commit history, CHANGELOG.md, TODOS.md, and various local project metadata files such as ~/.gstack/retro-context.md.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions for the processed content.
Capability inventory: The skill can execute shell commands via the Bash tool and write JSON snapshots to the project directory.
Sanitization: No explicit sanitization of commit messages or markdown content is defined before analysis.

retro