The Agent Skills Directory

[SAFE]: The skill implements strict environment validation by checking for the HERDR_ENV variable and verifying the availability of the herdr binary before execution.
[SAFE]: Checkout isolation is strictly enforced through mandatory git worktrees or dedicated branches, with explicit checks to ensure that destination directories are properly ignored and belong to the correct repository.
[SAFE]: The Zero attribution policy is consistently applied to maintain clean commit histories and reports, reflecting a specific user preference for output formatting.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external issue trackers, creating a surface for indirect prompt injection. 1. Ingestion points: Issues and PRDs are retrieved via gh api and Linear MCP tools in references/tracker-map.md. 2. Boundary markers: The skill uses structured templates in references/worker-prompt.md to isolate issue content from system instructions. 3. Capability inventory: The orchestrator uses herdr for tab management and gh/Linear tools for issue state updates. 4. Sanitization: Instructions explicitly restrict worker agents to the scope of a single issue and enforce identity verification before any code writes.

orchestrate-herdr