parallel-plan-executor CLAUDE.md

Repository instructions for parallel-plan-executor, a Claude Code workflow that runs independent implementation tasks at the same time.

In plain words
What is it for?
Use them when developing, testing, building, or reviewing the parallel-plan-executor workflow and its related Claude Code plugin.
Why use it?
They explain the project’s commands, architecture, branches, and dependency handling so contributors can change or test the repository consistently.

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/bacsystem/parallel-plan-executor/claude-md
Clone the repo
git clone --depth 1 https://github.com/bacsystem/parallel-plan-executor
Per session 1,072 This file is loaded in full into every session.
When invoked 1,072 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 $0.01072 $0.01072
Opus 5 $0.00536 $0.00536
Sonnet 5 $0.00214 $0.00214
Haiku 4.5 $0.00107 $0.00107

Measured yesterday against content hash 9cc2b53afd1c, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

parallel-plan-executor 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 · 40 lines

How it starts

The opening of the file, as written. The whole thing — 40 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Commands

npm test                            # run all unit tests (node --test)
node --test tests/scheduler.test.js # run a single test file
npm run build                       # regenerate workflows/parallel-plan-executor.js

Requires Node >= 20. ESM ("type": "module"), zero runtime dependencies.

What this repo is

A Claude Code Workflow script that executes a cys:plan implementation plan (### Task N: blocks with Consumes/Produces), running independent tasks in parallel instead of one at a time. The dependency DAG between tasks is inferred from each task's Consumes/Produces block and from files touched in common. The repo also ships the cys plugin (.claude-plugin/ + skills/): six skills covering design → plan → run → check → ship plus a guide index.

Architecture

The project is split in two halves because the Workflow sandbox has no filesystem access and forbids Date.now()/new Date():

  1. Local, pure-Node preparation (src/, bin/parse-plan.js) — runs outside the Workflow, fully unit-tested:

    • src/plan-parser.js parses ### Task N: blocks out of a plan file (normalizes CRLF first; only backtick-quoted symbols count in Consumes/Produces, matched one line at a time — a value wrapping to a second line is a known limitation, see README; duplicate task ids are rejected).
    • src/graph-builder.js infers dependencies (Produces→Consumes symbol matching, plus tasks touching the same file are chained — each depends on the last previous toucher) and rejects cycles.
    • bin/parse-plan.js ties both together and prints { tasks, graph } as JSON, which the user passes as the Workflow's args along with planPath, repoPath and integrationBranch (the branch every task merges into — required, so merge agents never guess).
  2. Workflow execution (workflows/):

    • workflows/parallel-plan-executor.js is generated — never edit it by hand. It is built by scripts/build-workflow.js, which inlines src/scheduler.js, src/graph-builder.js + src/validate-args.js, and src/time.js into workflows-src/parallel-plan-executor.template.js at the /* __SCHEDULER_SOURCE__ */, /* __VALIDATION_SOURCE__ */ and /* __TIME_SOURCE__ */ placeholders. The template lives in workflows-src/, not workflows/, so the plugin's skill auto-discovery never finds two files carrying the same meta block (Fase 4a fix — see docs/cys/specs/2026-07-17-fase-4a-quick-fixes-design.md). After changing any of those inlined modules or the template, run npm run build and commit both the source and the regenerated file.
    • Per task: implement in an isolated git worktree (branch task-<id>) → adversarial review (spec PASS/FAIL + quality APPROVED/NEEDS_FIXES) → at most one fix round → serialized merge (one at a time, DAG order). A failed task (including merge CONFLICT, BLOCKED, NEEDS_CONTEXT) cascades SKIP to all transitive dependents. After the final review, a Handoff agent prepares the git-flow closing (handoff.md with PR title/body and SemVer bump; with openPr: true it pushes and creates the PR — never merges it).
    • Wall-clock times come from the agents themselves (they run date); the script only does string math on HH:MM:SS values.

Read the full file on GitHub · 40 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 · 40 lines · 1,072 tokens per session scan A 9cc2b53afd1c

Subscribe to this mod's changes

parallel-plan-executor CLAUDE.md is an instructions file published in the GitHub repository bacsystem/parallel-plan-executor (2 stars, last pushed 1mo ago), licensed MIT. It adds 1,072 tokens to every session, about $0.0054 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.