The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from backlog item descriptions, git commit messages, and pull request metadata, which are then interpolated into prompts for various sub-agents (e.g., in swarm.md and plan.md).\n
Ingestion points: Data enters the agent context via the backlog view and backlog list tools, as well as git log and git show commands which retrieve repository history.\n
Boundary markers: The templates used for sub-agent prompts lack explicit delimiters or instructions to ignore instructions embedded in the external content.\n
Capability inventory: The skill has access to shell commands (git, uv, plan, backlog), the ability to modify the project backlog, and the capability to spawn powerful sub-agents like dh:add-new-feature and dh:implement-feature.\n
Sanitization: While the skill performs a 'Scope boundary check' in finalize.md to flag implementation-related keywords, it does not apply comprehensive sanitization to prevent the injection of instructions.\n- [EXTERNAL_DOWNLOADS]: The skill clones the 'Stateless Agent Methodology' repository from the author's official GitHub organization (bitflight-devops) to access methodology documentation.\n- [COMMAND_EXECUTION]: The skill extensively uses local CLI tools and scripts (git, uv, artifact, plan, backlog) to manage backlog items, track milestone progress, and coordinate planning tasks.

work-backlog-item