Farstax
OrganizationOpen-source, self-hosted runtime for Codex, Claude Code, Cursor, Antigravity/Agy and Grok Build. Keep durable coding-agent workstreams available from Telegram or Discord.
Categories
Indexed Skills (15)
autonomous-work
Execute a bounded autonomous Episode toward a durable Goal using ordinary Agent Bridge Runs, current evidence, and selective supervisor communication.
cli-auth-telegram
Use when a user chatting over Telegram asks to connect, authenticate, log in, or re-authenticate the Claude, Codex, Agy (Antigravity), or Grok Build CLI — walks them through the provider's OAuth/device flow with a clickable link and relays any code they paste back into the login process.
delivery-directives
Use when the user says "ship it", "review it", "release it", "deploy it", or "hotfix" to authorize the repository's existing delivery workflow without duplicating its engineering mechanics.
health-troubleshooting
Use when an Agent Bridge health observation needs evidence-led diagnosis, authorised remediation, recovery guidance, and verification
manage-mcp
Use when the user asks to add, configure, verify, update, troubleshoot, or remove an MCP server for Agent Bridge provider CLIs.
manage-skills
Use when the user asks to create, save, update, verify, repair, or remove a user-authored Skill without changing Agent Bridge's bundled Skills.
red-green-refactor-tdd
Use for software changes where behavior should be reproduced first, fixed with the smallest implementation, and protected by durable regression evidence.
release-readiness-review
Use for `review it` and before merge/release/deploy to adversarially test the intended outcome, real runtime journey, transitions, authority, and exact candidate evidence.
scheduled-routines
Align with the user on an explicit future or recurring instruction, then create, inspect, disable, or delete the agreed companion routine.
systematic-debugging
Use when investigating a bug, failure, or unexpected behaviour before proposing a fix
ui-engineering
Use for implementing or changing web UI where quality depends on design composition, rendered interface correctness, responsive layout, browser interaction, or visual verification rather than source and unit tests alone.
engineering-retro
Use when reviewing a recent body of engineering work, one release, or a set of releases to identify repeated human intervention and reusable friction worth fixing.
advisor
Get one bounded independent frontier opinion from a configured provider when the user asks for a second opinion or when an independent view would materially improve a decision.
git-sandbox
Use for isolating substantial changes or new features by creating git worktree feature branches, Draft PRs, and sandbox testing before merging.
requirements-to-acceptance
Use when turning product requests, bug reports, or ambiguous implementation asks into a small observable contract, affected boundaries, acceptance criteria, and verification plan.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.