The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and follow natural language instructions provided dynamically from the openspec tool's output via the context and operationGuidance fields.
Ingestion points: The agent parses JSON output from openspec instructions apply (SKILL.md, Step 3) and reads contents of contextFiles (SKILL.md, Step 4).
Boundary markers: The skill includes explicit instructions for the agent to "Keep both fields separate from CLI-returned state" and to "preserve the controlling value" in case of conflicts with built-in instructions (SKILL.md, Step 3).
Capability inventory: The skill possesses the capability to execute openspec CLI commands via Bash and perform file system modifications to implement tasks (SKILL.md, Step 6).
Sanitization: The skill implements logical sanitization by directing the agent to report conflicts and prioritize its core instruction set and CLI-controlled values over the external project-level guidance.
[COMMAND_EXECUTION]: The skill performs multiple shell operations using the openspec CLI for project management tasks (status, list, instructions, etc.).
Evidence: Multiple bash snippets in SKILL.md (Steps 2 and 3) execute the openspec command with various flags. The allowed-tools configuration restricts this execution scope to openspec:* subcommands, which helps mitigate broader command injection risks.

openspec-apply-change