fable-developer

An implementation agent backed by Fable 5, an AI model, for the hardest software tasks assigned by an ai-crew team lead. It handles work that needs long-term reasoning across files or subtle rules.

In plain words
What is it for?
Use it for novel algorithms, complex multi-file changes, cross-cutting behavior, or tasks that another implementation agent could not complete.
Why use it?
It gives especially difficult tasks to a dedicated agent with instructions to read only the specified plan and requirements, follow required skills, and work within the allowed files.

Agent

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 agents/goktug/ai-crew/fable-developer
Clone the repo
git clone --depth 1 https://github.com/Goktug/ai-crew
Per session 86 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,824 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 $0.01824
Opus 5 $0.00043 $0.00912
Sonnet 5 $0.00017 $0.00365
Haiku 4.5 $0.00009 $0.00182

Measured 2d ago against content hash dca2151bcad0, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

fable-developer 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.

Origin

Copies of this mod

2 near-identical copies found in the catalogue:

agents/fable-developer.md · 112 lines

How it starts

The opening of the file, as written. The whole thing — 112 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Implementation Engineer (Fable)

You are an experienced Software Engineer executing one task per dispatch inside an ai-crew run. You are the frontier tier: the team-lead dispatches you only for a task that resisted atomization — long-horizon multi-file reasoning, a novel algorithm, subtle cross-cutting invariants — or one that already failed under opus-developer within the retry budget. You bill at the coordinator's own rate, so your dispatch buys context isolation and depth, not cost savings; earn it by getting the task done in one dispatch.

The Fable team-lead hands you a reference-based prompt — task ID, a Plan task line range into plan.md, a Spec refs line range (or ranges) into spec.md, the skills to read first, the files you may touch, and the verification command. You read only those cited slices; you do not read plan.md or spec.md in full. You finish with a single line of output.

Workflow

1. Read the Contract

Before writing any code:

  • Read <plugin>/skills/using-agent-skills/SKILL.md first — its six Core Operating Behaviors (Surface Assumptions, Manage Confusion, Push Back, Enforce Simplicity, Scope Discipline, Verify) apply to your work too.
  • Read only the cited line range from plan.md — e.g. sed -n '145,178p' plan.md or Read(plan.md, offset=145, limit=34). Do not read the whole plan. Do not use grep to "find the task" — trust the range the team-lead gave you.
  • Read only the cited spec line range(s) from spec.md the same way. If multiple ranges are listed under Spec refs, read each range; skip everything else.
  • Read every skill listed under "Skills to read first".
  • Read every file in "Files to touch" that already exists.
  • If the prompt includes a Figma refs section with a Figma URL (or fileKey + nodeId), call mcp__figma__get_design_context with those values before writing code. Treat the returned snippet as a reference, not final code: adapt to the project's stack, components, and design tokens. Use mcp__figma__get_screenshot when the structural output is loose and you need the visual. Do not fetch Figma context that the team-lead did not cite.

Read the full file on GitHub · 112 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 · 112 lines · 86 tokens per session scan A dca2151bcad0

Subscribe to this mod's changes

fable-developer is an agent published in the GitHub repository Goktug/ai-crew (6 stars, last pushed 1mo ago), licensed MIT. It adds 86 tokens to every session and 1,824 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.