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 skills add AuraVixStudio/caelo --skill implementgit clone --depth 1 https://github.com/AuraVixStudio/caeloWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/auravixstudio/caelo/implement)<a href="https://agentmods.dev/skills/auravixstudio/caelo/implement"><img src="https://agentmods.dev/badge/skills/auravixstudio/caelo/implement/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/auravixstudio/caelo/implement"><img src="https://agentmods.dev/badge/skills/auravixstudio/caelo/implement.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What 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 5.1 | $0.00024 | $0.00443 |
| Opus 5.5 | $0.00010 | $0.00177 |
| Sonnet 5.5 | $0.00005 | $0.00089 |
| Haiku 4.5 | $0.00002 | $0.00044 |
Grade A, and why
Implement (multi-agent loop) 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 1mo 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.
What it actually says
Implement (multi-agent loop)
Use this when you (the orchestrator) should implement a non-trivial change by fanning
work out to subagents and iterating until the change is clean. You drive the loop with
the delegate tool; subagents work in isolated copies and their changes await the
user's merge review.
Loop
- Plan & split. Break the task into independent, self-contained chunks. Note the acceptance criteria you will check against.
- Implement.
delegateoneimplementerper chunk (parallel — pass them in a singledelegatecall). Give each a precise, standalone instruction. - Review. When the implementer summaries return,
delegateareviewer(one, or one per area) to review the changes for correctness, security and quality. Tell the reviewer exactly which files/areas to look at. - Fix. Collect the reviewer findings. If there are any blocking findings,
delegateanimplementerto fix them (reference the specific findings). - Repeat steps 3–4 until reviewers report no blocking findings, or you reach the round limit. Default to 3 rounds; if the user asked for an "effort N" (1–5), use N rounds.
- Summarize. Report what changed, the final review verdict, and that the changes are staged for the user's merge review.
Notes
- Keep your own context clean: integrate the returned summaries, not transcripts.
- Do not exceed the team limits (parallelism / subagent count / turn budget) — they are enforced; design your fan-out to fit.
- If a chunk turns out to depend on another, sequence those
delegatecalls instead of running them in parallel.
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.
- 1mo ago First seen · 38 lines · 24 tokens per session scan A 5f447fad7307
Implement (multi-agent loop) is a skill published in the GitHub repository AuraVixStudio/caelo (25 stars, last pushed 5d ago), licensed Apache-2.0. It adds 24 tokens to every session and 443 once invoked, about $0.0001 per session on Opus 5.5. 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 2026-08-30.
Other skills, from other repositories
api-reviewer
Use when the user asks to review or change an HTTP, RPC, event, CLI, or SDK API contract. Inspect routes, schemas, authorization, errors, compatibility, and consumers before editing, then verify request and response behavior with focused tests.
code-reviewer
A code-review guide for examining a change and its surrounding behavior for bugs, regressions, security risks, compatibility problems, and missing tests. A code review checks proposed changes before they are accepted.
pull-request-author
Use when the user asks to prepare, update, audit, or submit a pull request. Inspect repository contribution rules, branch state, diff, tests, and the target base first; create a focused reviewable change description and do not commit, push, or open a hosted pull request without explicit authorization.
refactor-engineer
Use when the user requests a behavior-preserving code refactor, cleanup, extraction, consolidation, or architectural simplification. Map all references and contracts before editing, keep the change scoped, and prove behavior with focused tests and type checks.
receiving-code-review
Handle review feedback on code with care: understand each point, check it against the real code before changing anything, fix what is right, and push back with reasons on what is wrong. Use when review comments arrive.
address-pr-comments
Work through the review comments on a pull request: check each point against the code, make the changes that are right, answer the ones that are not, and hold every reply for approval before it is posted.