The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection as it is designed to analyze local repository content and project files to determine integration parameters. 1. Ingestion points: The skill reads project manifest files (package.json, pyproject.toml, etc.), database schemas, environment variables, and source code via the triage.md and quickstart.md logic. 2. Boundary markers: There are no explicit markers or delimiters used when the agent processes these external data sources. 3. Capability inventory: The skill has access to Read, Grep, Glob, WebFetch, and AskUserQuestion tools. 4. Sanitization: The skill does not implement explicit sanitization or filtering for data ingested from the repository. While this poses a risk of the agent following instructions embedded in code comments or metadata, the skill enforces a human-in-the-loop policy and a 'Plan-first' contract to ensure the user reviews all proposed actions.
[EXTERNAL_DOWNLOADS]: The skill fetches technical documentation, integration presets, and installation guidelines from official Svix domains (docs.svix.com, api.svix.com, svix.com) and GitHub repositories. This behavior is consistent with the skill's purpose as a vendor-provided integration assistant and is used to ensure the agent uses the most current reference material.
[COMMAND_EXECUTION]: The skill provides detailed instructions for using the Svix CLI (svix) for tasks like local development relaying and shell scripting. It guides the agent in recommending correct command-line arguments to the user but does not autonomously execute system-level commands that would bypass user oversight.

svix-sending-webhooks