Borrowing it
Nothing to install: this file belongs to dean0x/devflow. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/dean0x/devflow/main/CLAUDE.mdgit clone --depth 1 https://github.com/dean0x/devflowWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/dean0x/devflow/claude-md)<a href="https://agentmods.dev/instructions/dean0x/devflow/claude-md"><img src="https://agentmods.dev/badge/instructions/dean0x/devflow/claude-md/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/instructions/dean0x/devflow/claude-md"><img src="https://agentmods.dev/badge/instructions/dean0x/devflow/claude-md.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.03208 | $0.03208 |
| Opus 5.5 | $0.01283 | $0.01283 |
| Sonnet 5.5 | $0.00642 | $0.00642 |
| Haiku 4.5 | $0.00321 | $0.00321 |
Grade A, and why
devflow 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 today.
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.
How it starts
The opening of the file, as written. The whole thing — 155 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Devflow Development Guide
Instructions for developers and AI agents working on Devflow. For user docs, see README.md.
Purpose
Devflow enhances Claude Code with intelligent development workflows. Modifications must:
- Maintain brutal honesty in review outputs (no sycophancy)
- Preserve context across sessions
- Enhance developer empowerment without replacing judgment
- Ensure all commands are self-documenting
Architecture Overview
Registry-driven CLI tool with 21 plugins (12 core + 9 optional). Plugins are entries in DEVFLOW_PLUGINS in src/core/plugins.ts — each entry declares its commands, agents, skills, and rules arrays. All assets live once in src/assets/; most install directly, and .mds sources compile via npm run build:mds — command hosts to dist/commands/, agent generator hosts to dist/agents/, and reference modules in src/assets/mds/ to dist/skills/git/references/.
Invariants:
- LLM-vs-plumbing: the LLM does all detection, matching and curation. Shell and TypeScript hold no detection or judgment logic.
- Read and write curation data only through the curation ops. Never edit those files. Never read the old
.devflow/learning/folder as a ledger. Only move it. - Ledger IDs are machine-local. Committed and posted text states a decision in words, never by ID.
- One install per machine, no install scope.
DEVFLOW_DIRis ignored.init --scope localexits 1. - devflow never writes
project.jsonorpolicy.json. The team commits them. - A git-tracked
.devflow/config.jsoncounts as absent, with a warning (D-PERSONAL-UNTRACKED). - Repo files only narrow the memory, curation and knowledge switches (D-FEATURES-NARROW-ONLY).
features.behavioris off unless set (D-BEHAVIOR-OPT-IN).- Prompts never read
project.jsonorconfig.json. They take theresolve-settings.cjsline (_partials/_settings.mds). - Learn
conventions.mdonly whenAPPLY_CONVENTIONSistrue. - Uninstall keeps
~/.devflow/tracker/{provider}.md. It is user content. - The Knowledge agent commits only
features/index.mdand{slug}/KNOWLEDGE.md(commit --only). Nevergit add -A, push or force. ensure-root-gitignoreandensureDevflowGitignorewrite the same bytes.KNOWLEDGEgates write-back only. Explore and debug do not load up-front. Freshness: write-through and verify-on-read only.- Rules install only for selected plugins and keep
paths: []frontmatter. - The
complianceskill and rule and thebehaviorrule are feature-owned. No plugin lists them. - Skills install for selected plugins and their
requires:. Never install or delete a dormant shadow. - The compliance lens is the union of all layers. A branch never removes a framework; an unreadable file never lowers it (D-LENS-UNION).
- The compliance directive goes into a spawn prompt only when the lens is on. No agent body names the compliance skill (D-COMPLIANCE-INJECT).
- Enabled compliance, or a repo
compliancekey (never HEAD), makesrequiredthe evidence floor. The lens never decides evidence. With compliance off no posting step is installed; underrequiredeach command prints one localTRACEABILITY: DEGRADEDline. - Flag
--enable/--disableare boolean-only. Narrow withisEnvBooleanFlag. Never display'unset'. - Disable and uninstall remove only devflow's exact value. Init adopts a user's own value as off (D-ADOPT-RESPECTS-USER).
TRACKER_MODELand the memory hook fallbacks equal their shipped rows.- Where compliance is on, PR-comment publication is visibility-gated and fails closed to a stub (D10).
project.jsononly caps it.requiredturnsoffintostub. - Where compliance is on, the test-plan comment makes no visibility probe: STUB under
auto, FULL underfull, none underoff. - Every posted body passes the scrubber (D11). If it fails, post nothing and emit
TRACEABILITY: DEGRADED (redaction unavailable). - File sinks chain the scrubber with
&&. Tool-call sinks use--emitand check the byte count [DR-06]. - Review-cycle convergence warnings never block.
- Evidence policy comes only from the default branch's
project.json(D-OFFLINE-ORIGIN-HEAD). Failures resolverequired. - Never parse
policy.json. Its presence resolvesrequired(D-POLICY-JSON-RETIRED). - Operations see only
ISSUE_REQUIRED,APPLY_CONVENTIONSandREQUIRE_NON_AUTHOR_APPROVAL. No prompt restates the mapping. - Only
pr-evidence.cjsandverify-evidence.cjsassign test-plan states. /code-reviewand/bug-analysiscarry no ticket gate.- Every install carries every tracker provider.
session-start-contextSection 3 always installs and gates at runtime. requires:lists used, unowned skills. The closure guard has one exception: Review's focus reference.- Tracker settings come from the settings line.
config.jsononly narrows them. - A tracker spawn reads
_contract.mdonce, and_mcp.mdonce off GitHub. A PR-host spawn reads neither. - Two or more qualifying servers: degraded run, no call.
- Release evidence matches the project key, never a bare number.
- A refusal reports
TRACEABILITY: DEGRADED (<reason>). On Linear, post the back-link with the dedup warning. tracker --setchanges only the machine default.- The Tracker agent claims first, then writes its conventions file once or not at all, scrub-gated.
- The manifest keeps the curation switch in
curationandlearning. Every write sets both. False wins. - When git tracks the curation or memory folder, or does not answer, that feature reads and writes no file there.
- The machine switch alone picks the posting variant; a repo key never restores it, and the lens stays the run-time union. One converge,
convergeVariantAxes, serves init,curationandcompliance. With compliance off it fails closed unless the prompts are proved free of posting.
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.
- today Changed · -1 lines · -1 tokens per session 6f3ba3a24078
- 6d ago Changed · -1 lines · -21 tokens per session 5fbfdd1bfff0
- 8d ago Changed · +1 lines · +8 tokens per session e9a4e581d4d7
- 14d ago Changed d8a49d6c0ca2
- 15d ago Changed · +23 lines · -2,081 tokens per session 09f0274b2278
- 16d ago Changed · +2 lines · +218 tokens per session 16d5b72af5bf
- 17d ago Changed · +2 lines · +311 tokens per session bbd3ebb21d1b
- 18d ago Changed · +63 tokens per session 5f28a958c3d7
- 19d ago Changed · -3 tokens per session bdeb008a775e
- 20d ago First seen · 129 lines · 4,714 tokens per session scan A b5142df01cd3
devflow CLAUDE.md is an instructions file published in the GitHub repository dean0x/devflow (21 stars, last pushed today), licensed MIT. It adds 3,208 tokens to every session, about $0.0128 per session on Opus 5.5. 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.
Other instructions, from other repositories
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, cache eligibility and rollout and applying the notice (cutover).
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
vscode copilot-instructions.md
Copilot instructions for microsoft/vscode, covering vs code copilot instructions, project overview, root folders, core architecture (src/ folder) and built-in extensions (extensions/ folder).
deepseek-harness AGENTS.md
AGENTS.md instructions for deepseek-ai/deepseek-harness, covering agents.md, pre-stable apis and released session data, repository layout, commands and host sandbox failures.