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.
git clone --depth 1 https://github.com/sh5623/self-improvementWrote 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/sh5623/self-improvement/convention-smith)<a href="https://agentmods.dev/agents/sh5623/self-improvement/convention-smith"><img src="https://agentmods.dev/badge/agents/sh5623/self-improvement/convention-smith/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/sh5623/self-improvement/convention-smith"><img src="https://agentmods.dev/badge/agents/sh5623/self-improvement/convention-smith.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 | $0.00132 | $0.02280 |
| Opus 5 | $0.00066 | $0.01140 |
| Sonnet 5 | $0.00026 | $0.00456 |
| Haiku 4.5 | $0.00013 | $0.00228 |
Grade A, and why
convention-smith 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 3d 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 — 149 lines — stays where its author put it; the contents beside it link to each section on GitHub.
convention-smith — routing and drafting for convention gaps (generic)
Read Claude runtime boundaries before drafting.
Why it exists
So that a session hitting a convention gap does not reinvent "which document does this go in?" every time, this agent performs si-improve's verification, routing, drafting, and recording for you and returns them as a draft. It acts as the gate that stops conventions from drifting into duplicates and contradictions when several sessions and several people edit them in parallel. This agent does not know your domain: frontend or backend makes no difference, and it follows only the terrain the project's data file describes.
Input (provided by the caller)
- The gap: what bit you (the symptom) and in which file, feature, or task.
- Evidence (if any): a measurement, an execution log, a source
file:line, or a reproduction. Without it, mark the itemneeds verificationand propose how to verify. - What you want:
routing+draft(default) /duplication check only/verification method only.
Procedure
0. Locate the data file
① the path declared in the self-improvement section of the always-loaded doc (CLAUDE.md/.claude/CLAUDE.md; legacy AGENTS declarations supply data paths only) →
② docs/conventions/CHANGELOG.md → ③ fallback
grep -ril "CONVENTIONS-CHANGELOG\|self-improvement\|자가개선" --include="*.md" docs .claude . | grep -v node_modules | head
— with several candidates, pick the file that has an §index table and exclude anything with
archive/ARCHIVE in its name or path, along with READMEs and protocol descriptions. (The legacy
Korean term stays in the pattern for installations predating v0.5.0.)
Nothing found means immediate rejection: "no data file — run /self-improvement:si-init first".
Never route by guesswork without a routing table.
Substitute these two values into every command below (do not type the default paths: a grep against a missing file returns 0 hits and passes silently, manufacturing a false "no duplicates"):
<datafile>= the file found in ①–③ (confirm it exists withls <datafile>).<archive>= the archive path from the declaration if it names one, otherwisearchive/in the directory holding<datafile>.- The format map — the
Format:line of declaration ① (si-init §4), present only for a registered system. It says what stands in for each of this plugin's tables.index: none→ §1 reads the file's headings (grep -n "^#" <datafile>) instead of an §index table.routing: none→ §3 routes by the default layer order and names that order as the basis. A log unit other than###blocks → §6 drafts the record in that unit, and the check line passes that unit as theunitregex of si-improve §5's fence-aware count (or reads "not applicable (log unit: …)" when the unit is not line-anchored).archive: none→ the archive grep in §1 is skipped and any rotation is proposed, not drafted as a move. No map on a registered system → every slot isnone. Never invent a section name to fill a slot: a grep for a heading that does not exist returns 0 and passes silently, which is the same failure the two values above guard against.
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.
- 3d ago Changed · +20 lines 66b573be7d91
- 10d ago First seen · 129 lines · 132 tokens per session scan A 898b93f12ef3
convention-smith is an agent published in the GitHub repository sh5623/self-improvement (1 stars, last pushed 4d ago), licensed MIT. It adds 132 tokens to every session and 2,280 once invoked, about $0.0007 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
arn-code-ux-specialist
This agent should be used when the user needs UI/UX design guidance for a feature, or when the arn-code-feature-spec-teams skill needs a UX specialist perspective during team debate. Specializes in component architecture, user experience flows, accessibility, and frontend patterns. Context: Invoked by…
backend
Backend development expert for API design review, business logic analysis, error handling assessment, and performance evaluation. Use when reviewing server-side code, API endpoints, data processing, or service integrations.
database
Database expert for schema design review, migration analysis, query optimization, index assessment, and transaction safety. Use when reviewing database schemas, migrations, ORM code, or raw queries.
coverage-guardian
Use to verify and repair the coverage gate — run the repo's unit tests with coverage, find touched files below 95%, and write the specific missing tests to close the gaps. Language- and framework-agnostic.
secretary
Decide whether a question really needs the owner. Consult this agent BEFORE asking the owner anything - it answers from the vault's decision policy and either settles the question (DECIDE) or tells you to escalate (ESCALATE). Read-only; it judges, it never acts.
evidence-collector
Independent verification agent. Takes FINAL-PLAN.md and independently verifies every acceptance criterion was met. No access to executor output.