The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it ingests data from external Linear issues and comments to drive internal workflows.
Ingestion points: According to references/lifecycle.md, the skill reads Linear issue titles, descriptions, comments, and links during the proposal binding phase.
Boundary markers: The instructions do not specify any delimiters or warnings to prevent the agent from following instructions potentially embedded within the Linear issue content.
Capability inventory: The skill is authorized to perform local file system writes, execute the openspec CLI tool, and perform network operations via the Linear MCP tools.
Sanitization: No content sanitization or validation is applied to data retrieved from Linear before it is used to populate repository artifacts or influence task generation.
[COMMAND_EXECUTION]: The skill invokes the openspec command-line utility to perform project lifecycle tasks.
Evidence: SKILL.md and references/lifecycle.md include instructions to execute openspec list --json, openspec validate <change> --strict, and openspec instructions apply.
Context: These commands are integral to the OpenSpec framework for listing, validating, and applying changes to the repository.
[DATA_EXFILTRATION]: The skill facilitates the transfer of repository-local specification data and proposal details to Linear.
Evidence: Instructions in SKILL.md direct the agent to mirror content from openspec/specs/ to Linear Project Documents and synchronize business-facing sections of proposal.md with Linear issue descriptions.
Context: This behavior represents the intended synchronization functionality between the repository and a well-known external project management service.

openspec-linearized