The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources (OpenSpec stores) that could potentially influence the agent's actions or lead to unauthorized file access.
Ingestion points: The skill parses JSON output from openspec status and openspec instructions, specifically the contextFiles array, dynamic instruction, context, and operationGuidance fields.
Boundary markers: The instructions provide logical guardrails, advising the agent to keep external context separate from CLI-returned state and to report conflicts rather than bypassing blocked states.
Capability inventory: The skill executes Bash commands (limited to the openspec CLI), reads arbitrary file paths provided in the contextFiles list, and writes to local files to update task checkboxes.
Sanitization: The skill relies on the agent's reasoning to identify conflicts rather than programmatic sanitization of the CLI output.
[COMMAND_EXECUTION]: The skill operates by executing system commands through the openspec CLI via Bash. Although the execution environment is scoped using allowed-tools, the skill's logic depends on the successful invocation and trusted output of these shell commands.

openspec-apply-change