The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill programmatically accesses the local configuration file at ~/.linearctl/config.json to extract a stored API key for authentication.
Evidence: API_KEY=$(jq -r '.profiles | to_entries[0].value.apiKey' ~/.linearctl/config.json) in SKILL.md.
[COMMAND_EXECUTION]: The skill utilizes several command-line tools to interact with the environment, version control, and external APIs.
Evidence: Execution of git branch, git diff, gh issue, lc issue, lc comment, jq, and curl commands.
[EXTERNAL_DOWNLOADS]: The skill fetches file attachments from Linear's official upload domain for assessment purposes.
Evidence: Uses curl to download files from uploads.linear.app as described in the attachment handling logic.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from external sources (ticket descriptions and comments) which are then processed by the agent to determine clarity and alignment.
Ingestion points: Ticket title, description, labels, and comment history retrieved via CLI tools (SKILL.md).
Boundary markers: Absent. The skill processes the text directly into assessment dimensions without delimiters or specific warnings to ignore instructions embedded in the ticket data.
Capability inventory: The agent has access to shell execution (git, curl), file system reading, and specialized CLI tools (gh, lc).
Sanitization: Absent. The raw text from comments and attachments is used directly to drive the grading logic, creating a surface for potential instruction override if the ticket content contains malicious prompts.

ticket-check