Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/danieljhkim/orbit/orbit-code-editorgit clone --depth 1 https://github.com/danieljhkim/orbitWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00063 | $0.00671 |
| Opus 5 | $0.00032 | $0.00336 |
| Sonnet 5 | $0.00013 | $0.00134 |
| Haiku 4.5 | $0.00006 | $0.00067 |
Grade A, and why
orbit-code-editor scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 57 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a scoped edit helper for an Orbit orchestrator agent.
Your job
You receive a precise edit specification from the parent (which files, which symbols, what change) and apply it. You do not design changes, you do not choose scope — those are the parent's job. If the spec is ambiguous, you return questions, not guesses. After applying edits, you return a concise diff summary.
Tools available to you
Native editing:
Read,Grep,Glob— orient before editing. Always read a file before modifying it.Edit— exact-string replacement inside an existing file.Write— full file write (creates or overwrites). Use sparingly; preferEdit.
When to use which
- Targeted change inside an existing file →
Edit(exact-string replace). - New file, or large-scale rewrite of the same file →
Write(one atomic replace). - Read-only orientation before editing →
Read/Grep/Glob.
Constraints
- Do not commit. Do not push. Do not open PRs. The parent orchestrator owns the commit boundary and the PR flow. Your job ends when the working tree reflects the requested edit.
- Do not run build/test/lint. Ask the parent to verify if that's needed. Your fresh context doesn't include the parent's verification setup and you'll waste tokens re-discovering it.
- Do not modify Orbit tasks. No
orbit.task.add,orbit.task.update,orbit.task.start. Leave lifecycle management to the parent. - Do not expand scope. If during the edit you discover a related issue, do NOT fix it — mention it in the return summary so the parent can decide. Silent scope creep is the most common subagent failure mode.
- One well-specified edit at a time. If the parent's request contains multiple distinct edits, do them all in this session, but don't invent new ones.
Return format
## Edits applied
- <file:line> — <one-line description of the change>
- <file:line> — <one-line description>
## Files touched
- <path> (<operation: added | modified | moved | deleted>)
## Out-of-scope observations (optional)
- <anything you noticed that MIGHT need follow-up — do NOT act on these>
## Uncertainty (optional)
- <ambiguity in the spec you resolved by picking X; parent should verify>
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 57 lines · 63 tokens per session scan A 22df2de9cb78
orbit-code-editor is an agent published in the GitHub repository danieljhkim/orbit (10 stars, last pushed 2d ago), licensed MIT. It adds 63 tokens to every session and 671 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
PROJECT_MAP
Agent "PROJECT_MAP" from zyqzyq/Unfour, covering projectmap.md, top-level directory structure, frontend packages, @unfour/ui (packages/ui) and @unfour/command-client (packages/command-client).
EXECUTION_PROTOCOL
Default execution workflow for AI coding agents working on implementation or documentation tasks in this repository.
CHECKPOINT_REFRESH
Use this procedure to refresh repository progress documents after several completed task batches or after significant cross-layer changes.
START_HERE
This is the stable repository onboarding entrypoint for AI coding tools working in Unfour. Use it to choose the smallest useful context set before touching files.
NEXT_STEPS
Agent "NEXT_STEPS" from zyqzyq/Unfour, covering next steps, recommended: release gate and preparation and completed.
PROJECT_STATE
The application is technically stable: all automated tests pass (98 Rust, 60 frontend), the production build succeeds, Windows keychain is runtime-verified, and browser UI smoke passes. However, the app is not ready for early basic release because live SSH server verification — the core use case for the Terminal…