si-improve

si-improve is a skill for Claude Code from sh5623/self-improvement. It costs 48 tokens per session (303 once invoked), scanned A, original, MIT.

A procedure for improving project conventions when missing or unclear instructions cause problems during work. It finds the relevant convention record, checks the evidence, and documents a reusable fix.

In plain words
What is it for?
Use it after discovering a recurring process or documentation problem. It helps verify the issue, decide where the rule belongs, draft the update, and record it.
Why use it?
It prevents the same instruction gap from causing repeated confusion across files or tasks. It also helps avoid adding duplicate or contradictory rules.

Skill for Claude Code

Written for Claude Code: shipped in a Claude Code plugin. Also seen: mentions Codex; $skill-name invocation.

Part of the self-improvement plugin — 3 skills, 1 agent, 1 hook shipped together

Good fit Use it after discovering a recurring process or documentation problem. It helps verify the issue, decide where the rule belongs, draft the update, and record it.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/sh5623/self-improvement/si-improve
Install

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.

Any agent
npx skills add sh5623/self-improvement --skill si-improve
Clone the repo
git clone --depth 1 https://github.com/sh5623/self-improvement

Made for: Claude Code.

Or install self-improvement, the plugin that ships this one along with the rest of its 3 skills, 1 agent, 1 hook.

Wrote 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.

agentmods badge for si-improve

README.md
[![agentmods](https://agentmods.dev/badge/skills/sh5623/self-improvement/si-improve/github.svg)](https://agentmods.dev/skills/sh5623/self-improvement/si-improve)
Your own site
<a href="https://agentmods.dev/skills/sh5623/self-improvement/si-improve"><img src="https://agentmods.dev/badge/skills/sh5623/self-improvement/si-improve/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.

agentmods 80×15 button for si-improve

Your own site · 80×15
<a href="https://agentmods.dev/skills/sh5623/self-improvement/si-improve"><img src="https://agentmods.dev/badge/skills/sh5623/self-improvement/si-improve.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 48 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 303 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00048 $0.00303
Opus 5 $0.00024 $0.00151
Sonnet 5 $0.00010 $0.00061
Haiku 4.5 $0.00005 $0.00030

Measured 4d ago against content hash 702883e8f5c2, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-11, from the pricing page.

Security

Grade A, and why

si-improve 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 4d 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.

plugins/self-improvement/skills/si-improve/SKILL.md · 28 lines

What it actually says

Repair a convention in Codex

Read runtime boundaries, data resolution, and the repair protocol.

Resolve the real project data file and format map once. If none exists, use $self-improvement:si-init within the authorized work. Treat Claude-only instructions as belonging to Claude, while reusing the shared data, evidence and canonical project rules.

Apply the protocol in order: gate → classify → verify → codify → effectiveness check → propagate → record. Name recurrence sites and evidence before writing a global rule. Inspect the project tool configuration before choosing prose. Preserve frozen artifacts, existing authorization, log format, budgets and other sessions' changes.

For uncertain routing or drafting, use the convention-smith reference. It is usable in the main session; independent delegation is optional and requires an available, authorized agent. Claude's named agent registration is not a dependency.

At the end, report the adopted codifications and real paths, verification, remaining propagation targets, and the project's self-improvement marker. Rejected or local-only candidates do not count as codifications.

Changes

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.

  1. 4d ago Changed · -198 lines · -84 tokens per session 702883e8f5c2
  2. 11d ago First seen · 226 lines · 132 tokens per session scan A 7716fde24149

Subscribe to this mod's changes

si-improve is a skill published in the GitHub repository sh5623/self-improvement (1 stars, last pushed 5d ago), licensed MIT. It adds 48 tokens to every session and 303 once invoked, about $0.0002 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.

Related

Other skills, from other repositories

align

Align a project to brana practices — assess gaps, plan, implement, verify. Auto-detects type. Use when setting up a new project or realigning an existing one.

martineserios/thebrana · 39 tokens

recursive-spine-digest

Use when sweeping all recursive-spine-conforming repos for tracking health — aging deferrals (oldest first), stalled milestones, in-flight by lane, assigned-but-untouched issues, and debts named in closing comments but never filed, plus seedling pollen that never transplanted (hive repos). Reports honest denominators…

slopstopper/recursive-spine · 82 tokens

recursive-spine-scaffold

Use when a repo has (or is getting) the tracking stamp and needs the rest of its spine — interviews for and stamps up to six parts, each optional: a rules codex with a moments map, an ADR directory, a CI gate skeleton, a session-memory convention, a constraints file with its sha-pinned drift gate, and the loop…

slopstopper/recursive-spine · 0 tokens

recursive-spine-bootstrap

Use when installing the recursive-spine tracking convention onto a repo — interviews for modules and dialect, then stamps labels, issue/PR templates, a CLAUDE.md/AGENTS.md tracking section, and cross-project board membership. Idempotent; degrades loudly on missing gh scopes; offers (never forces) plumb-line and…

slopstopper/recursive-spine · 76 tokens

recursive-spine-nudge

Use after a recursive-spine-digest sweep to select and deliver system-initiated nudges — 0–3 conversation-starters pushed to the owner about what is unblocked, aging, stalled, or newly connected. Every nudge ends in a question; the skill proposes conversations, never executes work.

slopstopper/recursive-spine · 68 tokens

recursive-spine-pollinate

Use when something just proved itself in real work and should be carried to other projects — captures it as a pollen record in the builder's hive (registry file + paired issue), or, in pull mode, reads every configured hive and offers relevant transplants into the current repo, recording each transplant back onto the…

slopstopper/recursive-spine · 91 tokens