Home / turbokach / claude-code-setup · skills/analyze-arcs/SKILL.md · GitHub

analyze-arcs skillA

analyze-arcs is agent-read markdown (skill) from turbokach/claude-code-setup: Measure the Claude Code pipeline sessions since a date — master transcripts, their subagents/*.jsonl, and the codex-challenge logs — and report the numbers plus the mechanical doctrine violations. Invoke on "analyze the sessions", "how did the last arcs go", "evaluate the pipeline", "/analyze-arcs <since-date>", or before any kit change that claims to fix a pipeline problem..

Indexed from public GitHub and served as immutable, content-addressed versions. Install it pinned to an exact SHA-256 with the mdr CLI, and every file is verified against the hash recorded here before it reaches your agent. The deterministic audit below grades the latest version, and the same file always earns the same grade.

What the file says

# Analyze arcs

The scripts are read-only. Times in the report are UTC; transcripts are under
`~/.claude/projects/<project>/<session>.jsonl` with subagent transcripts beside them in
`<session>/subagents/`. The master transcript is a partial record: owner↔subagent messages
and everything a subagent did live only in the subagent files, so never judge an agent from
the master's view of it.

## Procedure

1. **Version diff first** (global rule: optimize starts with a version diff). `claude --version`
   against the stamp in the kit's `docs/references.md`; if they differ, read the changelog
   between them before blaming the doctrine for a harness change.
2. **Run the measurement** into the session scratchpad:
   ```
   python3 ~/.claude/skills/analyze-arcs/scripts/analyze.py --since YYYY-MM-DD --out <scratchpad>/arcs/report.md
   ```
   Per session: span, models, Claude Code version, peak context, subagent roster (type, pin, turns, minutes, KB,
   model actually seen), every `codex-challenge.sh` launch with its range, `--out`, run
   minutes and verdict size, every gate with how long it waited, pushes, killed background
   tasks. Then the **flags** — mechanical checks only:
…

Read the whole file at its exact version.

How to install

Latest version
mdr add turbokach/claude-code-setup/analyze-arcs@git:20260920.b04ab81
Exact content
mdr add turbokach/claude-code-setup/analyze-arcs@sha256:c0f2037f695e7f00

Pin to a label to follow the author's releases, or to a sha256 to freeze the exact bytes forever. Either way the resolved hash is written to mdr.lock, and mdr install reproduces it on any machine.

Badge

mdr badge

[![mdr](https://markdownregistry.com/badge/art_kdxlwofkb4jkdoku.svg)](https://markdownregistry.com/a/art_kdxlwofkb4jkdoku)

1 badge views in 30 days

Versions

versioncommittedcommitsizeaudit
git:20260920.b04ab81 latest2026-09-20 b04ab81 7,110 BA view · diff
git:20260918.f2b0b852026-09-18 f2b0b85 6,827 BA view · diff
git:20260914.1ee72182026-09-14 1ee7218 6,790 BA view · diff
git:20260913.aad644a2026-09-13 aad644a 6,769 BA view · diff
git:20260913.b526a662026-09-13 b526a66 6,077 BA view

Audit of the latest version

A  17 of 17 checks passed. Deterministic, no model, same answer every run.
  • pass: Frontmatter block present
  • pass: Frontmatter declares a name
  • pass: Frontmatter declares a description
  • pass: Size between 200 bytes and 200 KB (7110 bytes)
  • pass: No zero-width or bidi control characters
  • pass: No instruction hidden inside an HTML comment
  • pass: No link to an exfiltration or paste host
  • pass: No credential-shaped string
  • pass: No instruction to send local credentials anywhere
  • pass: No text hidden with inline styles
  • pass: No prompt-injection phrasing
  • pass: No curl or wget piped into a shell
  • pass: No recursive delete of root, home or parent
  • pass: No instruction to read or print local credentials
  • pass: No base64 blob over 200 characters
  • pass: No link to a raw IP address
  • pass: No script tag

Source

GitHub

turbokach/claude-code-setup · 10 stars · license MIT · pushed 2026-09-23 · branch master

API

GET https://markdownregistry.com/api/v1/artifacts/art_kdxlwofkb4jkdoku
GET https://markdownregistry.com/api/v1/resolve?ref=turbokach/claude-code-setup/analyze-arcs
GET https://markdownregistry.com/api/v1/blob/c0f2037f695e7f009a372ca132bd0f56aa2f691a0d9d48ac0fb91585ccecff63

Your agent does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.

More from turbokach/claude-code-setup

CLAUDE.md@global claude
turbokach/claude-code-setup · global/CLAUDE.md
git:20260923.1c604ea · audit A · 10 stars
agent-teams skill
turbokach/claude-code-setup · skills/agent-teams/SKILL.md · Orchestration playbook for parallel multi-agent work in Claude Code. Use when fanning out genuinely parallel…
git:20260923.28280ab · audit A · 10 stars
feature-workflow skill
turbokach/claude-code-setup · skills/feature-workflow/SKILL.md · The user's six-stage feature pipeline (discuss → plan in native plan mode → validate + approve → delegated execute →…
git:20260923.7ddafe9 · audit A · 10 stars
stack-update skill
turbokach/claude-code-setup · skills/stack-update/SKILL.md · Apply a pending claude-code-setup update — fetches upstream, shows what changed, three-way-diffs CLAUDE.md against your…
git:20260819.fced18d · audit A · 10 stars

Every file in turbokach/claude-code-setup

Browse by kind, by grade A, or by owner.