The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute local CLI tools including canon (for database record management), orchestra (for project state visualization and analysis), and pytest (for verifying task implementation).
Evidence: Commands like orchestra analyze, canon apply, and python3 -m pytest are used throughout SKILL.md and the reference files.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data—such as specifications, design reviews, and handoff documents—to generate actionable task and acceptance records, creating a surface for indirect prompt injection.
Ingestion points: The agent is instructed to read active specifications, spec/design reviews, decisions, and constraints from the project repository (SKILL.md Step 1, references/planning-mechanics.md Step 1).
Boundary markers: No explicit delimiters or boundary instructions are defined for the input data; however, the skill mandates structured output in NDJSON format for the canon tool.
Capability inventory: The skill can execute shell commands, write to a local SQLite database via canon, and trigger subagent tasks (SKILL.md, references/oracle-author.md).
Sanitization: The risk is mitigated by a mandatory human review gate (plan-review question) that requires a user to approve, change, or reject the generated plan before any execution occurs (SKILL.md Step 7).

tasks