The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run within its Python script to execute the prettier formatting tool on generated Markdown files. It also calls the qmd utility to update and embed vault indexes. These operations are restricted to formatting and indexing tasks and include pre-execution environment checks.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to the official Linear GraphQL API (api.linear.app) to fetch project and ticket data. These operations are limited to read-only queries and target a well-known, trusted service domain.
[DATA_EXFILTRATION]: The skill accesses the local file path ~/.config/linear/api-key and environment variables to retrieve the Linear API token. This is a standard method for managing service credentials for CLI-based tools. The script includes specific logic to ensure the key is never logged or interpolated into error messages.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data (ticket titles, descriptions, and comments) from Linear. It mitigates potential injection risks through several layers:
Ingestion points: Data is fetched from the Linear API and processed by the scripts/linear_sync.py script rather than the agent directly.
Boundary markers: Machine-generated content is strictly delimited by  and  markers to protect user-owned notes.
Capability inventory: The skill's write capabilities are limited to a specific vault subdirectory (01 Projects/Linear/), and it explicitly lacks Edit or Write tool permissions for the agent.
Sanitization: The render_markdown function performs extensive sanitization, including stripping unauthorized HTML, demoting headings to preserve note structure, and masking code blocks during the rewriting process.

obsidian-linear-sync