The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on shell command execution for its core functionality, specifically interacting with the Git version control system and the 'orchestra' and 'canon' workflow tools.
Evidence: Frequent use of git branch, git worktree, git merge, and git commit to manage task-specific environments.
Evidence: Orchestration commands such as orchestra wave start, orchestra task land, and orchestra verify are used to drive the task lifecycle.
Evidence: Database interactions performed via canon update, canon list, and canon create to maintain project state.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from task briefs and subagent evidence reports, creating a surface for indirect prompt injection.
Ingestion points: Data enters the parent agent's context from orchestra brief output and the JSON evidence blocks returned by subagents.
Boundary markers: The skill attempts to mitigate risks by using JSON for subagent returns and enforcing isolated write roots for subagent work.
Capability inventory: The agent can execute arbitrary shell commands and perform filesystem/database writes, which could be abused if the ingestion process is compromised.
Sanitization: No explicit validation or sanitization is performed on the content of subagent observations or task briefs before processing.

execute