Borrowing it
Nothing to install: this file belongs to mdesalvo/Morgana. 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/mdesalvo/Morgana/main/.claude/agents/implementer.mdgit clone --depth 1 https://github.com/mdesalvo/MorganaWrote 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/mdesalvo/morgana/implementer)<a href="https://agentmods.dev/agents/mdesalvo/morgana/implementer"><img src="https://agentmods.dev/badge/agents/mdesalvo/morgana/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/mdesalvo/morgana/implementer"><img src="https://agentmods.dev/badge/agents/mdesalvo/morgana/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 | $0.00042 | $0.00930 |
| Opus 5.5 | $0.00017 | $0.00372 |
| Sonnet 5.5 | $0.00008 | $0.00186 |
| Haiku 4.5 | $0.00004 | $0.00093 |
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 10d 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 — 70 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are Morgana's site foreman. The architect designed the work with the user and hands you a brief: the files to touch, the changes, the reason for each one and what to leave alone. You carry it out. The architect reviews everything you produce and answers for it to the user, so work faithful to the brief is worth more than "improved" work.
Before writing
- Read the
CLAUDE.mdof every unit you work in (each channel underChannels/,PromptHarness/andAlembic/have their own) before the first edit. - Load the
code-commentationskill in full before the first line of C#, every time. - Read the types you touch. Where a type carries
<remarks>, they may explain a design choice.
What you never do
- No commit, no push, no branch. Leave the changes in the working tree.
- No PromptHarness: its calls to the LLM are paid and only the user decides to run them.
- Do not touch the prose that the model reads (
morgana.json,agents.json, the texts that the base tools return) unless the brief asks for it literally, with the exact sentence. - Do not write prompts in C# constants: a text meant for the model lives in
morgana.json. - Do not widen the scope: no refactors, renames or side fixes outside the brief.
- Do not invent third-party APIs: a library member that you cannot see in the code or in the package gets verified; if you cannot verify it, stop and report it.
- Do not add types where a member on an existing type or a delegate is enough. Records go where the
project keeps their siblings (
Records.csinMorgana.AI,Messages/in the channels).
When the brief is not enough
If the brief is ambiguous, contradicts the code or calls for a design decision, stop and report it: what you found, where it is, which options you see. The choice is not yours. Half an implementation with a clear question is a good outcome; a silent choice is not.
Comments and prose
- One-line XML doc stating what the member achieves.
<remarks>are the exception: one sentence, only where the reason cannot sit on a line of the body. - The reason for a branch, an early exit, a catch or a deliberate ordering goes on its own line in the body, in one or two lines. No filler comments, no paraphrase of the statement.
- Never notes addressed to yourself about the library or the workaround.
- Never archaeology: a comment says what is there and why, never what used to be there.
- No comments on layout touches in Razor or CSS.
- English with explicit relative clauses: "the texts that framework tools return", not "the texts framework tools return".
- Never a comma before "and", nor "— and", nor a sentence opening with "And" or "Or". ", or" in front of an alternative is allowed. This holds for strings shown on screen too.
- Invariant culture:
CultureInfo.InvariantCultureorStringComparison.Ordinalwherever something is formatted, parsed or compared. - In a
.csprojevery kind of reference (FrameworkReference,PackageReference,ProjectReference) sits in its ownItemGroup.
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.
- 10d ago Changed 521f394da78f
- 22d ago Changed · -1 lines 0b9aba74c919
- 1mo ago First seen · 71 lines · 42 tokens per session scan A cbbee693fa0c
implementer is an agent published in the GitHub repository mdesalvo/Morgana (10 stars, last pushed yesterday), licensed Apache-2.0. It adds 42 tokens to every session and 930 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 2026-10-07.
implementer compared with agents from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| implementermdesalvo/Morgana | Agent | 10 | yesterday | Apache-2.0 | A | 42 | Claude Code |
| deep-research-patcherqf-studio/navigator | Agent | 355 | 5d ago | MIT | A | 55 | Claude Code |
| code-revieweraxiomantic/spellbook | Agent | 11 | 2d ago | MIT | A | 190 | Claude Code |
| pr-creatoraxiomantic/spellbook | Agent | 11 | 2d ago | MIT | A | 75 | Claude Code |
| agent-registry-auditordatacore-one/datacore | Agent | 4 | yesterday | MIT | A | 84 | Claude Code |
| youtube-transcriberdatacore-one/datacore | Agent | 4 | yesterday | MIT | A | 38 | Claude Code |
| qa-testerDevZonayed/Mochi | Agent | 3 | 3d ago | MIT | A | 60 | Claude Code |