The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of vendor-provided CLI tools (comet, comet state, comet guard) to perform workspace management, state initialization, and artifact validation. These commands are executed as local subprocesses.
[EXTERNAL_DOWNLOADS]: Instructions specify loading external sub-skills such as openspec-explore and openspec-new-change via a platform Skill tool. The skill includes instructions to adapt and constrain these external commands using a local adapter (comet classic openspec).
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect injection as it processes untrusted user-provided content. * Ingestion points: User-provided PRDs, roadmaps, and product proposals are read and analyzed in SKILL.md and reference/batch-changes.md. * Boundary markers: There are no explicit instructions to use delimiters or ignore instructions embedded within the user-provided documents. * Capability inventory: The agent has the ability to execute shell commands and write to the file system. * Sanitization: Names derived from the input are normalized to kebab-case, but the document content is used to populate artifact templates without further sanitization.
[DYNAMIC_EXECUTION]: The agent is directed to execute commands based on argv and cwd values provided dynamically in the JSON output of the comet CLI tool during the artifact generation phase. This creates a runtime dependency on external tool output for command construction.

comet-open