The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by ingesting and acting upon untrusted data from the local repository and tool outputs.
Ingestion points: The skill reads delta spec files identified in artifactPaths.specs.existingOutputPaths and processes context and operationGuidance from the openspec instructions command output.
Boundary markers: The skill contains explicit instructions to keep these external inputs separate from controlling logic (conflict analysis, user choices, CLI checks) and to ignore or report conflicts when external guidance contradicts hard constraints.
Capability inventory: The skill utilizes the Bash tool to execute openspec commands, performs file system operations including directory creation (mkdir) and moving files (mv), and triggers the openspec-sync-specs workflow.
Sanitization: The agent is instructed not to copy external text verbatim into specs and to prioritize its own codebase investigation and CLI-derived data over advisory instructions.

openspec-bulk-archive-change