The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. It retrieves issue titles and bodies from external sources (Linear API and GitHub CLI) and interpolates this untrusted data directly into prompts for sub-agents using the assets/implementer-template.md template.
Ingestion points: Phase 1 uses linear api and gh issue list to fetch issue content (SKILL.md).
Boundary markers: The assets/implementer-template.md lacks explicit boundary markers or instructions to the sub-agent to ignore instructions embedded within the issue body.
Capability inventory: The skill executes local scripts via go run (Phase 0, 0.5, 1, 2, 3, 5), performs git operations (git worktree add, git commit), and dispatches sub-agents (Phase 2).
Sanitization: No evidence of sanitization or validation of the retrieved issue content before it is used to drive automated workflows.
[COMMAND_EXECUTION]: The skill frequently executes local Go scripts using the pattern go run ../../scripts/cmd/<name>. These scripts are used for critical operations including state assessment, event emission, contract creation, and data persistence. While the scripts are local to the project, their execution is driven by parameters derived from external issue data.

product-implement