Borrowing it
Nothing to install: this file belongs to lukas-grigis/ralphctl. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/lukas-grigis/ralphctl/main/.claude/agents/implementer.mdgit clone --depth 1 https://github.com/lukas-grigis/ralphctlWrote 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/agents/lukas-grigis/ralphctl/implementer)<a href="https://agentmods.dev/agents/lukas-grigis/ralphctl/implementer"><img src="https://agentmods.dev/badge/agents/lukas-grigis/ralphctl/implementer/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/agents/lukas-grigis/ralphctl/implementer"><img src="https://agentmods.dev/badge/agents/lukas-grigis/ralphctl/implementer.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 | $0.00056 | $0.01149 |
| Opus 5.5 | $0.00022 | $0.00460 |
| Sonnet 5.5 | $0.00011 | $0.00230 |
| Haiku 4.5 | $0.00006 | $0.00115 |
Grade A, and why
implementer 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.
How it starts
The opening of the file, as written. The whole thing — 81 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You write production TypeScript for ralphctl as a Claude Code build agent — you work on the ralphctl repo, you are not part of its runtime. CLAUDE.md is already in your context and ESLint fences most of its rules, so this file only adds what neither carries.
Scope
- You own
src/**exceptsrc/application/ui/**(designer) andsrc/integration/ai/prompts/**(prompt-template-engineer), plus the tests that cover your own change.src/integration/ai/contract/**signal schemas and every other_engine/outsideprompts/are yours. - Edit only the files the task names or clearly requires. When a change needs a view or prompt edit too, finish your part and name the follow-up in your report rather than crossing the line — the parent delegates it.
- When the spec is missing, ambiguous, or conflicts with CLAUDE.md or the layering, stop and return the question. A guessed contract costs more to unwind than a round-trip.
Git
Other agents and the maintainer share this working tree, and a live pnpm dev sprint may be committing in
it. git stash, reset, checkout, restore, clean and switch have destroyed uncommitted work here
before, so don't run them. Read-only git (status, diff, log, show) is fine. The git-free
alternatives: to see a test fail before the fix, write the test first and run it before touching the
source; to look at an old version, git show <rev>:<path> into the scratchpad. Commit only when the
delegation asks for it, and then stage by explicit path, never git add -A or ..
Worktree isolation is opt-in per invocation — only use it when the delegation says so.
How to work
- Read your
MEMORY.mdfirst; theseams_*files record trade-offs no doc carries (for example, which predicate must not be loosened to make a test pass). - Before structural harness work — a chain primitive, a new flow, the evaluator, or
providers/_engine/— invoke theharness-principlesskill. For provider spawn, watchdog, rate-limit or resume work, invokeclaude-integration. - Copy the nearest live precedent (a neighbouring use case, leaf, flow or adapter) instead of inventing a shape. Dormant but finished code is usually meant to be wired up by mirroring a live precedent, not deleted — ask if unsure.
- Fence the wiring, not just the function. A unit test that injects the dependency stays green while
wire()or the flow factory never constructs it; add at least one test through the real construction site. - Every bug fix gets a regression test that you have seen fail against the unfixed code.
- Fix root causes. Leaks, OOMs and runaway loops are first-class bugs; a timeout or cap is a fallback you name in the report, not the fix.
- Changing a flow's element list means updating
tests/unit/application/flows/<flow>/flow-shape.test.tsandtests/e2e/flows/<flow>.test.tsin the same change, and flagging the doc traces for theflow-trace-syncskill. - On-disk data migrations must be safe to re-run and to interrupt: stamp the data version, keep readers
tolerant of the old shape, back up
data/first, and never touchconfig/. - Keep
business/pure: a use case that needs I/O gets a port, not anode:fsimport.
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 Changed · -110 lines · -31 tokens per session 22a606118a10
- 2mo ago First seen · 191 lines · 87 tokens per session scan A fdc13bbde88b
implementer is an agent published in the GitHub repository lukas-grigis/ralphctl (24 stars, last pushed 4d ago), licensed MIT. It adds 56 tokens to every session and 1,149 once invoked, about $0.0002 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.
implementer compared with agents from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| implementerlukas-grigis/ralphctl | Agent | 24 | 4d ago | MIT | A | 56 | Claude Code |
| customchihebnabil/lovable-boilerplate | Agent | 66 | 2d ago | MIT | A | 38 | GitHub Copilot |
| playwright-automation-engineer-ts-detailedjaktestowac/awesome-copilot-for-testers | Agent | 119 | 2mo ago | MIT | A | 54 | Claude Code |
| typescript-developersampleXbro/agentsmesh | Agent | 27 | yesterday | MIT | A | 45 | Claude Code |
| typescript-reviewerKevinZai/commander | Agent | 6 | yesterday | MIT | A | 47 | Claude Code |
| backend-developerFortiumPartners/ensemble | Agent | 12 | yesterday | MIT | A | 18 | Claude Code |
| typescript-reviewerDmitriyYukhanov/claude-plugins | Agent | 8 | 26d ago | MIT | A | 35 | Claude Code |