dev-agent CLAUDE.md

A guide for developing the `dev-agent` Claude Code plugin, including its commands, agents, session files, and 15-stage implementation workflow.

In plain words
What is it for?
Use it when changing the plugin, understanding `/implement` behavior, or resuming work tracked in `.devagent/` files.
Why use it?
It gives coding agents the repository-specific rules needed to modify the plugin without confusing its source files with the files that run in target projects.

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/surjal/dev-agent/claude-md
Clone the repo
git clone --depth 1 https://github.com/Surjal/dev-agent
Per session 1,868 This file is loaded in full into every session.
When invoked 1,868 The same file — it is already loaded in full.
Security scan C 1 finding. 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 $0.01868 $0.01868
Opus 5 $0.00934 $0.00934
Sonnet 5 $0.00374 $0.00374
Haiku 4.5 $0.00187 $0.00187

Measured 2d ago against content hash bc34dc897275, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade C, and why

dev-agent CLAUDE.md scanned grade C with 1 finding 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.

Recursive force deletehighDestructive command

rm -rf with a variable or a broad path is one typo away from removing the wrong tree.

- Never delete data or run destructive commands (`rm -rf`, `git reset --hard`, `git push --force`,
CLAUDE.md · 121 lines

How it starts

The opening of the file, as written. The whole thing — 121 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md — dev-agent plugin source repo

This repo IS the source of the dev-agent Claude Code plugin. The actual runtime behavior that ships to target projects lives in agents/*.md and commands/*.md at the repo root (plugin convention — see .claude-plugin/plugin.json) — those files must stay self-contained, since a target project has no access to this file.

This CLAUDE.md only governs sessions working ON this repo (developing the plugin itself).

Workflow the plugin implements (see commands/implement.md for the authoritative version)

Since v1.4.0, every /implement run is "Deep Mode" (the only mode — no separate fast path) and follows the same numbered 15-stage lifecycle; commands/implement.md → Execution stages has the authoritative table. .devagent/ inside the target project persists progress (state.json, plan.md, progress.md, decisions.md, failures.md) so /implement --resume can pick a session back up — see docs/deep-execution.md. What follows are the same task-shape → stage-selection patterns as before, now expressed as which of the 15 stages actually run:

Default (bug fix / small change — unchanged since before architect/ux-designer/ frontend-developer existed):

Understand → Research (researcher) → Plan → Implement (developer) → Test (tester)
   → Review (reviewer) → Definition of Done → PASS+APPROVED → done
                        → FAIL / CHANGES REQUIRED → back to developer → Test again

Frontend feature, Playwright available in the target project:

Research (researcher) → UX Designer → Frontend Developer → Test (tester) → Visual QA (visual-qa)
   → Review (reviewer) → Definition of Done → PASS+APPROVED → done
   → Visual QA FAIL → back to Frontend Developer → Test → Visual QA again

Frontend feature, Playwright unavailable: same as above minus Visual QA, reported explicitly as "Visual QA skipped: browser automation capability unavailable" rather than silently omitted.

Full pipeline (new project / major feature — "One-Shot Project Builder"):

Read the full file on GitHub · 121 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 · 121 lines · 1,868 tokens per session scan C bc34dc897275

Subscribe to this mod's changes

dev-agent CLAUDE.md is an instructions file published in the GitHub repository Surjal/dev-agent (2 stars, last pushed 17d ago), licensed MIT. It adds 1,868 tokens to every session, about $0.0093 per session on Opus 5. A static security scan graded it C with 1 finding (recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.