The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided mission briefs and transforms them into executable code and test suites, creating a surface for indirect prompt injection.\n
Ingestion points: Primary data enters via brief.md and flows into spec.md, plan.md, and implementation steps.\n
Boundary markers: The skill lacks explicit prompt delimiters or "ignore previous instructions" markers in its templates to isolate untrusted brief content from agent instructions.\n
Capability inventory: The skill possesses significant capabilities, including writing files to the repository, executing shell commands for testing, and managing external workflow state via CLI tools.\n
Sanitization: No explicit sanitization or validation of the mission brief content is described before processing.\n- [EXTERNAL_DOWNLOADS]: The skill references several external repositories and framework presets to define its operational lanes.\n
Evidence: It links to repositories under tikalk, obra, mattpocock, and addyosmani to provide framework-specific logic for specifying, planning, and coding.\n
Note: Resources from the author's organization (tikalk) are vendor-owned. Other references target well-known frameworks and developers in the agentic development community.\n- [COMMAND_EXECUTION]: The skill uses command-line tools to manage its lifecycle and execute local scripts.\n
Evidence: It invokes adlc-cli for workflow management and runs local scripts such as run-branch.sh and open-mr.sh to manage git operations and pull requests.\n- [DYNAMIC_EXECUTION]: The skill dynamically executes tests and implementation logic generated during the mission lifecycle.\n
Evidence: The test step executes commands to verify a failing state (RED gate), and the code step involves running the generated suite to verify completion (GREEN gate).

mission-brief