prd-taskmaster CLAUDE.md

Instructions for Atlas and Claude Code, describing a system where a leading AI model divides work among command-line agents and direct APIs. The guide also covers Atlas's task marketplace, protocol, and completion checks.

In plain words
What is it for?
Use it when changing Atlas's driver, task routing, executor marketplace, protocol integration, backend model, or checks for whether work is done.
Why use it?
It keeps changes consistent with Atlas's goal of routing each task to an appropriate executor while preserving evidence that the result is reachable and correct.

Instructions file

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 instructions/anombyte93/prd-taskmaster/claude-md
Clone the repo
git clone --depth 1 https://github.com/anombyte93/prd-taskmaster
Per session 2,637 This file is loaded in full into every session.
When invoked 2,637 The same file — it is already loaded in full.
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 $0.02637 $0.02637
Opus 5 $0.01319 $0.01319
Sonnet 5 $0.00527 $0.00527
Haiku 4.5 $0.00264 $0.00264

Measured yesterday against content hash 212d7192deb1, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

prd-taskmaster CLAUDE.md 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 yesterday.

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.

CLAUDE.md · 158 lines

How it starts

The opening of the file, as written. The whole thing — 158 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Claude Code Instructions — Atlas

The end goal (north star — test every change against this)

Atlas's entry point is a frontier model (Fable, or whatever the best frontier model is) acting as the driver. The driver understands the user's intent using skills built for it, then — inside a loop — decomposes the goal into optimal components and routes each to the cheapest capable executor, to save as many tokens as possible while delivering an outcome as good as if the frontier model had done all the work itself, at a fraction of the cost.

The executor pool is any CLI agent harness (Claude Code, Codex, Gemini, whatever ships next) plus direct API paths (the native backend's parse/expand/rate), all in parallel. Not a fixed provider list. Every feature decision is tested against: does this make the frontier-driver-decomposes-to-cheap-executors loop better, cheaper, or more provable?

What makes Atlas a product, not a prompt pack (the four moats):

  1. Cross-vendor arbitrage — route work across any harness/API by cost and capability.
  2. Engine-enforced unfakable gates — validate-tasks, ship-check (SHIP_CHECK_OK), the two-axis "done" (oracle + reachability, below), structural tests. If a feature could be a skill, it's not a product feature until the engine enforces it.
  3. Persistent vendor-neutral task graph — tasks.json survives across vendors/sessions.
  4. Cross-vendor cost ledger — telemetry + economy-report prove the savings.

Unfakable "done" — the two gates (5.3.0+, the trust backbone)

A task ships only when both axes pass — done = oracle-PASS AND code WIRED/EXEMPT. This is the property the whole product (and the marketplace below) rests on; do not weaken it.

  • Gate 5 — re-execution oracle. The binding ship gate is skel/ship-check.py (oracle-backed). It re-runs the operator-held tests at the claimed commit in a network-isolated, digest-pinned podman sandbox, so a submitter cannot pass by editing their own logs. prd_taskmaster/oracle_bridge.py maps a CDD card → graded verdict via the spine's atlas oracle CLI; it is fail-closed (any ambiguity → FAIL, never PASS). The self-grantable SHIP_CHECK_OVERRIDE_ADMIN bypass was removed. (prd_taskmaster/shipcheck.py is a NON-binding display heuristic only — never add an oracle call there.)
  • Gate 6 — reachability (prd_taskmaster/reachability.py, reachability-sweep CLI). A green test on an orphan module (imported by nothing) is blocked and surfaced as ⚠ scaffolded. Verdicts: WIRED (a non-test file imports it) / EXEMPT (declared cli:|route:|tool:|hook:|plugin:|dynamic: scheme, accepted on trust in v1) / ORPHAN (fail). Read-only at runtime.

Read the full file on GitHub · 158 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. yesterday First seen · 158 lines · 2,637 tokens per session scan A 212d7192deb1

Subscribe to this mod's changes

prd-taskmaster CLAUDE.md is an instructions file published in the GitHub repository anombyte93/prd-taskmaster (592 stars, last pushed 18d ago), licensed MIT. It adds 2,637 tokens to every session, about $0.0132 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 2026-08-30.