The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources including openspec instructions archive output, openspec instructions specs output, and local project files such as tasks.md and spec.md. This information is used as "context," "operationGuidance," and "rules" to influence the agent's behavior during the archive and sync process.
Ingestion points: Step 1 and Step 4 in SKILL.md describe fetching instructions and rules from CLI outputs and reading local markdown files.
Boundary markers: The instructions include several protective directives, such as "Keep both fields separate from built-in steps," "report the conflict and preserve the controlling value," and "Never copy runtime context, operation guidance, or artifact-rule text verbatim into output files."
Capability inventory: The skill utilizes tool capabilities for file system modification (mkdir, mv), reading/writing project files (intelligent spec merge), and executing bash commands via the openspec CLI.
Sanitization: The skill relies on the agent to manually identify and report conflicts between provided guidance and controlling inputs, rather than using automated sanitization or strict schema enforcement for the natural language instructions.
[COMMAND_EXECUTION]: The skill performs shell operations (mkdir -p, mv) in Step 5 using directory paths (changeRoot, planningHome.changesDir) that are dynamically parsed from the JSON output of the openspec status command.
[DYNAMIC_EXECUTION]: In Step 4, the skill performs an "agent-driven intelligent merge" to synchronize delta specifications into main specification files. This involves the agent generating new content for spec.md files based on external source data and dynamically retrieved merge rules.

openspec-archive-change