The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest external context and operation guidance from the local environment, which could be manipulated by an attacker to include malicious instructions.
Ingestion points: The skill reads data from openspec instructions archive, openspec instructions specs, and the tasks.md file in the SKILL.md (Steps 1, 3, and 4).
Boundary markers: The skill includes explicit instructions to treat external data as advisory and forbids it from overriding core workflow steps or command contracts. It also requires the agent to report conflicts.
Capability inventory: The skill can execute shell commands for filesystem management (mkdir, mv) and invoke the openspec CLI tool.
Sanitization: The skill relies on natural language boundary logic rather than programmatic escaping or validation of the fetched instructions.
[COMMAND_EXECUTION]: The skill executes shell commands to archive files and interact with the OpenSpec CLI.
Evidence: Use of mkdir -p and mv to manage the archive directory structure in Step 5.
Evidence: Multiple sub-process calls to the openspec CLI (e.g., status, list, instructions) with parameters derived from user input or context.

openspec-archive-change