The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple sources including openspec instructions archive (JSON), openspec instructions specs (JSON), and tasks.md (Markdown). These inputs provide 'context', 'rules', and 'operationGuidance' that influence the agent's behavior during the archiving and syncing process.
Ingestion points: Step 1 fetches archive context; Step 3 reads local task files; Step 4 fetches spec synchronization rules.
Boundary markers: The skill contains explicit instructions to treat these external inputs as advisory, report conflicts with internal logic, and not copy content verbatim into output files.
Capability inventory: The skill can execute commands via openspec, create directories with mkdir, and move files with mv.
Sanitization: The agent is instructed to consider guidance entry-by-entry and explain any inapplicable or conflicting advice rather than blindly following it.
[COMMAND_EXECUTION]: The skill executes shell commands restricted to the openspec CLI toolset and standard filesystem utilities. It uses mkdir -p to prepare archive directories and mv to relocate the change root to the archive destination. The commands are constructed using variables derived from the openspec status JSON output.

openspec-archive-change