The Agent Skills Directory

[PROMPT_INJECTION] (HIGH): Indirect Prompt Injection surface. Ingestion points: conductor/tracks/{trackId}/plan.md (task descriptions) and spec.md (requirements). Boundary markers: Absent; the agent is instructed to parse and follow the descriptions directly as its workflow loop. Capability inventory: Full file system write access for implementation and shell command execution (npm test, pytest, git). Sanitization: Absent; there are no instructions to validate or escape content from the plan files before execution or implementation.
[COMMAND_EXECUTION] (HIGH): Arbitrary command execution via test runners. The skill invokes shell commands such as npm test or pytest and potentially other verification tasks defined in the project's metadata. A malicious actor could embed shell-breaking characters or direct command payloads into the plan.md tasks or verification rules, which the agent would then execute with its current shell privileges.

conductor-implement