The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting and acting upon external, potentially untrusted data.
Ingestion points: The skill executes openspec instructions archive to retrieve context and operationGuidance, and openspec instructions specs to retrieve merge rules. It also parses tasks.md and the output of openspec status --json.
Boundary markers: The instructions include explicit guardrails, such as 'Keep both fields separate from built-in steps', 'do not copy their text verbatim', and 'report the conflict and preserve the controlling value'. These markers are intended to prevent the agent from being fully overridden by the external instructions.
Capability inventory: The agent has the capability to execute shell commands (via the openspec tool), create directories (mkdir), move directories (mv), and perform an 'agent-driven intelligent merge' of specification files.
Sanitization: While the skill instructs the agent to 'read and consider' inputs without copying them verbatim, the agent still integrates this external logic into its operational workflow, which can be exploited if the external source is compromised.
[COMMAND_EXECUTION]: The skill frequently executes shell commands to interact with the filesystem and the openspec CLI.
The skill uses mkdir -p to create archive directories and mv to relocate the change root directory.
It heavily relies on the openspec tool for operations like list, status, instructions, and archive.

openspec-archive-change