The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the repository, including project-specific orchestration files and GitHub issue/PR content, to coordinate complex workflows.
Ingestion points: The skill reads ORCHESTRATION.md, issue descriptions, and pull request metadata from the target repository.
Boundary markers: The skill architecture mandates the use of "context-isolated task packets" and "isolated" context modes for subagents to prevent instruction bleed.
Capability inventory: The skill possesses the ability to spawn subagents (/deliver), create issues (/create-issue), and publish releases (/create-release).
Sanitization: It implements "lane-admission preflights" to validate environmental constraints and a schema-validated event ledger to audit every material transition.
[COMMAND_EXECUTION]: The skill utilizes a bundled Python script for telemetry management and includes a corresponding test suite that executes shell commands.
Evidence: The workflow invokes python3 "$SKILL_DIR/scripts/event_ledger.py" for event ingestion and validation.
Evidence: The test file tests/test_event_ledger.py uses subprocess.run to verify the script's functionality.
Context: These executions are restricted to the skill's own deterministic logic and standard development testing practices.

milestone-rush