fable-review

A review process that asks fresh reviewers to challenge a plan, answer, design, analysis, or risky code change. It uses normal automated checks first, then focuses reviewers on mistakes those checks may miss.

In plain words
What is it for?
Use it for adversarial reviews, second opinions, red-team checks, and higher-risk changes; for pure code diffs, a dedicated code-review command may be more suitable.
Why use it?
It reduces the risk of trusting your own assumptions or overlooking important edge cases before work is accepted or shipped.

Skill for Claude CodeCodex

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.

agentmods
npx agentmods add skills/debabsah/fable-method/fable-review
Any agent
npx skills add debabsah/fable-method --skill fable-review
Clone the repo
git clone --depth 1 https://github.com/debabsah/fable-method

Made for: Claude Code, Codex.

Per session 86 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,016 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
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 $0.00086 $0.01016
Opus 5 $0.00043 $0.00508
Sonnet 5 $0.00017 $0.00203
Haiku 4.5 $0.00009 $0.00102

Measured 2d ago against content hash d0c488924cc2, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

fable-review 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 2d 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.

skills/fable-review/SKILL.md · 25 lines

How it starts

The opening of the file, as written. The whole thing — 25 lines — stays where its author put it; the contents beside it link to each section on GitHub.

fable-review

Don't self-vibe-check. Manufacture blind adversaries — fresh context, none of your rationale — and let them attack the actual work. Uses only the built-in Agent/Task tool — no other plugin required. (For a pure code diff, a dedicated code-review command in your environment may be sharper for that narrow case; this runner needs nothing installed and covers every artifact — designs, plans, analyses, schemas, configs, prose.)

Run it

  1. Run the deterministic checks first — the acceptance oracle, tests, linters, validators. Don't spend a reviewer on what a tool catches; reviewers are for judgment. Between the two sit different-kind checks — types, property/invariant checks, a reference implementation, a dry run on real data — more independent of your blind spots than another model instance; prefer one where it exists.
  2. Size the panel to the risk tier (the method skill's table): T2 → one lens, the dominant risk; T3 → 2–5 lenses. The default first lens is the scope block's load-bearing unknowns — attack what changes everything if wrong; generic lenses (correctness/logic, security, data/edge-cases, architecture, requirements-fit, ops) fill the rest. Going past the tier minimum needs a named reason — a specific unresolved risk, not thoroughness for its own sake. One concern per lens; no overlap.
  3. Dispatch one fable-lens subagent per lens, all in a single message (multiple Agent/Task calls in one turn = they run in parallel). fable-lens ships with this plugin (agents/fable-lens.md); its toolset is a harness-enforced allowlist (Read, Grep, Glob), so a reviewer cannot edit the artifact under review and cannot dispatch further subagents. That is a boundary, not a request — the difference matters, because "READ-ONLY" addressed to an agent holding Edit/Write/Bash is only a suggestion. Give each dispatch just its lens and the exact scope (files, diff range, or artifact + the plan/requirements it's judged against); the adversary rules live in the agent.

Read the full file on GitHub · 25 lines

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. 2d ago First seen · 25 lines · 86 tokens per session scan A d0c488924cc2

Subscribe to this mod's changes

fable-review is a skill published in the GitHub repository debabsah/fable-method (1 stars, last pushed 1mo ago), licensed MIT. It adds 86 tokens to every session and 1,016 once invoked, about $0.0004 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.