Skip to content
Agentmart
Submit a tool

cold-read

Load for /vellum:cold-read, after every major editorial program, and before export.

Skill by FiredMosquito831

IndexedApache-2.0★ 2
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/firedmosquito831-cold-read

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Load for /vellum:cold-read, after every major editorial program, and before export. The cold-read skill owns the cover-to-cover cold-read protocol: the reader charter, the rolling ledger, the append-only issue log, and the batching/orchestration model. It finds the accumulating-memory defect classes (clock drift, cross-chapter state contradictions, prop custody, promise bookkeeping, knowledge-state slips) that no scanner, chapter reviewer, or continuity query can see.

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/cold-read

What's inside

Skill name
cold-read
Description
Load for /vellum:cold-read, after every major editorial program, and before export. The cold-read skill owns the cover-to-cover cold-read protocol: the reader charter, the rolling ledger, the append-only issue log, and the batching/orchestration model. It finds the accumulating-memory defect classes (clock drift, cross-chapter state contradictions, prop custody, promise bookkeeping, knowledge-state slips) that no scanner, chapter reviewer, or continuity query can see.
License
Apache-2.0
Instructions
38 lines in SKILL.md · view on GitHub

Files (5)

  • SKILL.md
  • batching.md · resources/
  • charter.md · resources/
  • issue-format.md · resources/
  • reader-ledger.md · resources/

SKILL.md (first lines)

# Cold Read

The cold-read skill owns the cover-to-cover cold-read protocol — a single reader, reading the whole manuscript in order, carrying a rolling ledger, and logging every defect that survives the per-chapter QA. It exists because a class of defect is invisible to every memoryless pass: a linter sees a line, a chapter reviewer sees a chapter, a continuity query sees a query — none of them accumulate reader-state. Clock drift, cross-chapter state contradictions, prop custody, promise bookkeeping at book scale, knowledge-state slips, and audit coverage bias are all *accumulating-memory* defects. The cold read is the only pass that catches them.

## What this skill is

- A **protocol**, not a single prompt. It runs on four files (charter, ledger, issues, batch reports) that together make a multi-sitting read behave like one continuous reader.
- **Assessment-only by default** (rule zero): the read measures; it does not edit the thing it measures. AUTO-FIX is opt-in, narrowly scoped, logged.
- **Batched** for sustainability and context-window safety: 8–11 chapters per sitting, the ledger the only inter-sitting state.
- A **self-contained** reference: it restates the severity taxonomy, the issue format, the batching model, and the orchestration pattern so the agent and the muse can run it without loading other skills. Cross-references to `gates/`, `kb-integrity/`, and `story-ledgers/` are pointers only.

## The document set

| File | Role |
|---|---|
| `work/cold-reads/<YYYY-MM>/charter.md` | Who the reader is, the rules, the batch plan, the severity taxonomy, autonomy tiers. Written once. |
| `work/cold-reads/<YYYY-MM>/reader_ledger.md` | The reader's accumulated memory: NEXT marker, position/clock, promise register, knowledge map, prop custody, do-not-re-explain, PROTECT, tripwires, grades. Updated every batch. |
| `work/cold-reads/<YYYY-MM>/issues.md` | Append-only issue log, one line per issue, fixed format. |
| `work/cold-reads/<YYYY-MM>/batches/batch_X.md` | Per-batch report: what works, grades, seam assessment, "would a paying reader keep going?" |

## Load the resource needed

- `resources/charter.md` — the reader persona ("devoted genre reader who just finished the previous book, with a line-editor's ear; experience first, diagnosis second"), rule zero, the batch plan, the severity taxonomy, AUTO-FIX vs PROPOSE tiers, voice cards, tripwires, PROTECT, deliverables.
- `resources/reader-ledger.md` — the rolling ledger: NEXT marker with WATCH list, position/clock table, promise register (HOT/WARM/COLD), character knowledge map, prop custody, do-not-re-explain register, PROTECT, tripwires, open threads, grades table. "A stale ledger is an amnesiac reader."
- `resources/issue-format.md` — the fixed line format `CR-### | ch:line | SEV | CAT | "quote" | reader-moment | fix direction | SCOPE`, the severity/category vocab, the severity→shared-schema mapping, and the rules (reader-moment mandatory, append-only, continuous numbering).
- `resources/batching.md` — 8–11 chapter batches cut at natural milestones, the per-sitting resume recipe (charter + ledger + last ~40 issues → continue from NEXT), the per-batch procedure, the batch report contents, the orchestration model, and when to run.

## Who runs it

- **Muse** sets up `work/cold-reads/<YYYY-MM>/` from `templates/cold-read/` (via `/vellum:cold-read`), seeds the ledger, and spawns `@cold-reader` per batch.
- **`@cold-reader`** executes one batch: reads charter + ledger + last ~40 issues, continues from NEXT, reads only the batch's chapters, returns (a) append-ready issue lines, (b) ledger section updates, (c) the batch report. Assessment-only; no peeking at outline or editorial notes during a batch.
- For **15+ chapters** muse MAY use the digest pattern from `fiction/agents/review-coordinator.md` (MIT) — spawn per-batch readers, aggregate from summaries, cache per-batch reports on disk.

## How it fits the gates

Cold-read findings are advisory by default; **BLOCKER-class findings block export** until resolved or dismissed-with-author-signoff (the same discipline as `story-ledgers/` exemptions). MAJOR/MODERATE need a triage note. The export readiness checklist (`gates/resources/readiness.md`) cross-checks the cold-read `issues.md` for open BLOCKERs.

From FiredMosquito831/my-writing-companion, Apache-2.0.

Signals

Stars
2
Forks
0
Open issues
0
Last commit (this folder)
Sep 9, 2026 (1 month ago)
Last release
Sep 11, 2026 (1 month ago)

From GitHub (fetched Oct 24, 2026).

Details

License
Apache-2.0
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from FiredMosquito831

Cite this page

cold-read skill. Agent Mart. https://agentmart.dev/skills/firedmosquito831-cold-read (last updated 2026-10-24)