The Agent Skills Directory

[COMMAND_EXECUTION]: The skill automates complex workflows by executing numerous shell commands.
Evidence: Invokes project management and orchestration tools bd and ao throughout SKILL.md.
Evidence: Executes standard development tools including git, grep, jq, go, npm, and pytest.
Evidence: Performs build verification and executes project-generated binaries (e.g., ./bin/$BINARY --help) in Step 4a of SKILL.md.
[PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection via issue data processing.
Ingestion points: Processes issue IDs and full descriptions from external trackers or user input in SKILL.md.
Boundary markers: Absent; untrusted data is directly embedded into command strings and skill arguments.
Capability inventory: Includes the ability to write/edit files, execute arbitrary shell commands, and delegate tasks to other high-capability skills like test, review, and refactor.
Sanitization: No sanitization logic is present for data interpolated into commands such as ao context assemble --task='<issue title and description>'.

implement