The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the Bash tool to perform git operations (fetch, log, diff), manage local session files in ~/.gstack/, and interact with the gh and glab CLIs. It also modifies the CLAUDE.md file to inject routing rules.
[DYNAMIC_EXECUTION]: In the preamble and several steps, the skill uses eval and shell sourcing (source <(...)) on the output of local binaries located in ~/.claude/skills/gstack/bin/ (e.g., gstack-slug, gstack-repo-mode). These binaries are provided by the skill vendor.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests data from external, untrusted sources.
Ingestion points: Reads git commit messages, PR/MR titles, CHANGELOG files, TODOS.md, and a user-authored retro-context.md file.
Boundary markers: None identified. The data is interpolated directly into the analysis narrative.
Capability inventory: The agent has access to Bash (shell execution), Write (filesystem modification), and Read tools.
Sanitization: There is no evidence of sanitization or escaping of the ingested git metadata before it is processed by the model for narrative generation.

retro