codeflow: Instructions file for Claude Code

CLAUDE.md

codeflow CLAUDE.md is an instructions file for Claude Code, Codex from xiaojiou176-open/codeflow. It costs 6,065 tokens per session, scanned D, original, MIT.

A set of project instructions for working with a CLAUDE.md file, a document that tells coding tools how to understand and modify a code repository.

In plain words
What is it for?
Use it when maintaining the listed codeflow repository and its AI entrypoints, documentation, commands, generated files and local or hosted checks.
Why use it?
It gives the agent a defined reading order and working rules, helping changes stay small, verifiable and consistent with the project's documentation and security checks.

Instructions file for Claude CodeCodex

Written for Codex and Claude Code: reads ~/.codex or $CODEX_HOME, but also the file is CLAUDE.md. Also seen: mentions CLAUDE.md; mentions Claude Code; mentions AGENTS.md.

This is xiaojiou176-open/codeflow's own configuration. It tells Claude Code and Codex how to work on codeflow itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything codeflow configures →

Reuse

Borrowing it

Nothing to install: this file belongs to xiaojiou176-open/codeflow. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/xiaojiou176-open/codeflow/main/CLAUDE.md
Clone the repo
git clone --depth 1 https://github.com/xiaojiou176-open/codeflow

Made for: Claude Code, Codex.

Wrote 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.

agentmods badge for codeflow CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/xiaojiou176-open/codeflow/claude-md.svg)](https://agentmods.dev/instructions/xiaojiou176-open/codeflow/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/xiaojiou176-open/codeflow/claude-md"><img src="https://agentmods.dev/badge/instructions/xiaojiou176-open/codeflow/claude-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 6,065 This file is loaded in full into every session.
When invoked 6,065 The same file — it is already loaded in full.
Security scan D 2 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 5.1 $0.06065 $0.06065
Opus 5 $0.03033 $0.03033
Sonnet 5 $0.01213 $0.01213
Haiku 4.5 $0.00607 $0.00607

Measured 6d ago against content hash 9d1badedb1e6, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade D, and why

codeflow CLAUDE.md scanned grade D with 2 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 6d 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.

Reads agent configuration directoriesmediumAgent snooping

.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.

live-provider rule that resolves process env first and `~/.codex/config.toml`

Recursive force deletehighDestructive command

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

before the clean-room `rm -rf` sweep, plus quarantining stubborn dashboard
CLAUDE.md · 417 lines

How it starts

The opening of the file, as written. The whole thing — 417 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md

This file mirrors the root AI entrypoint for tools that prefer CLAUDE.md.

Read First

  1. README.md
  2. docs/index.html
  3. configs/repo_positioning.json
  4. configs/docs_nav_registry.json
  5. AGENTS.md

Working Rules

  • keep diffs small and auditable
  • run real verification commands before claiming success
  • keep docs and code synchronized
  • keep runtime output out of tracked source
  • anchor root AI entrypoints to public routes, config manifests, and receipts instead of archived internal markdown under docs/
  • keep public test/probe fixtures free of maintainer-local absolute paths and raw token-shaped literals; prefer generic workspace roots plus synthetic string assembly, keep placeholder security URIs pinned to the exact example.com contract, preserve .jsonl temp-report hints in portable scan scratch names, fail closed on tracked direct email/phone markers plus forbidden runtime files, fail closed on open GitHub secret/code scanning alerts in local hooks and pre-push while GitHub-hosted trusted_pr, untrusted_pr, and hosted-first push_main Quick Feedback / hosted policy lanes stay advisory under integration-token and first-analysis timing limits, keep workflow static security (actionlint + zizmor), canonical secret scanning, and Trivy dependency scanning wired into repo-owned entrypoints, and sync the root/docs entrypoints when that contract shifts
  • host-process safety is fail-closed: worker/test/orchestrator paths must not use killall, pkill, process-group kills, negative/zero PID signals, or AppleScript System Events; terminate only recorded child handles and stop with manual cleanup instructions when stale repo-owned runtime state already exists
  • repo-owned scripts/*.py entrypoints must keep shared helper imports usable when they are executed directly or through bash scripts/run_governance_py.sh; do not assume the repo root is already present on PYTHONPATH
  • host-compatible pre-commit hooks that execute repo-owned scripts/*.py entrypoints must use the same wrapper path (or an equivalent python3 -B contract) so clean hook runs do not leave repo-local __pycache__ residue
  • keep public CI hosted-first: fork PRs stay low-privilege on GitHub-hosted lanes, and sensitive verification stays on protected manual dispatch lanes
  • GitHub repo-first pushes may hand CI an all-zero base SHA; repo-owned doc-drift/doc-sync gates must skip ci-diff comparison in that case instead of failing before the repository has a real baseline commit
  • treat configs/github_control_plane_policy.json as the machine SSOT for required check names, and point human-facing summaries back to the root README.md instead of duplicating the list here
  • GitHub Control Plane workflow should prefer the repo secret GH_ADMIN_TOKEN when it needs to prove admin-only repository APIs, because the default workflow token cannot read Actions permissions, branch protection, or vulnerability-alert endpoints on the live control plane
  • protected upstream/live-smoke receipts are route-exempt on trusted_pr, untrusted_pr, and hosted-first push_main; those routes must not fail solely because manual closeout/provider credentials are absent
  • hosted push_main governance closeout also treats upstream_report, upstream_same_run_report, and current_run_consistency as advisory when the manifest already marks upstream/live smoke route-exempt
  • when dashboard dependency lock refreshes land, keep the app-local dashboard lockfile aligned with the root workspace lock updates and document the change
  • the current security-only dashboard lock refresh keeps [email protected] pinned through the repo-owned override layer so [email protected] does not drift back onto the vulnerable [email protected] transitive path, without expanding this maintenance slice into a broader Lighthouse upgrade
  • when dashboard or desktop lock maintenance changes the shipped dependency contract, update the root docs entrypoints in the same patch so doc-sync and closeout gates describe the live state
  • when one closeout patch touches both dashboard and desktop packaging, mirror that decision in the root docs entrypoints instead of relying on module docs alone
  • when the live public GitHub surface moves or changes repository URLs, sync the root docs/security/storefront entrypoints in the same patch so repo-side links do not drift behind the published Codeflow surface
  • when security reporting wording changes, keep SECURITY.md, SUPPORT.md, issue template contact links, and the root README aligned in the same patch
  • when security-scan or fixture-hygiene changes touch tracked test literals or scan wrappers, keep this file, AGENTS.md, and the root/docs entrypoints aligned in the same patch; current examples include generic workspace roots instead of maintainer-local absolute paths, runtime-built token-like fixtures, and BSD-safe temp-file naming in scripts/security_scan.sh
  • when dashboard/operator wording or intake/runtime contracts change, sync the root AI/docs entrypoints in the same patch so doc-sync gates keep following the live English-first dashboard surface and the current intake/probe rules
  • when runtime-provider compatibility changes the orchestrator client contract, sync the root AI/docs entrypoints in the same patch; current examples include the Switchyard runtime-first /v1/runtime/invoke adapter, the forced chat_completions mode on chat-only intake/operator paths, and the fail-closed rule that keeps MCP tool execution on tool-capable providers; Quick Feedback-safe helper extraction, dead-code-clean provider_resolution compatibility exports, and env-governance allowlist updates for read-only runtime-capability summaries follow the same rule
  • when role-contract / prompt-ref / handoff-summary semantics change the orchestrator contract or preview surfaces, sync the root AI/docs entrypoints in the same patch; current examples include resolved role_contract, intake role_contract_summary, summary/risk-only handoff, and the governance-backed metadata in policies/agent_registry.json plus configs/env_direct_read_allowlist.json
  • when Prompt 4-style binding/read-surface work extends those role-contract surfaces, keep the root AI/docs entrypoints aligned in the same patch; current examples include contract-derived role_binding_summary in PM-facing run_intake(...) responses plus the same summary persisted into run manifests, alongside registry-backed SEARCHER/RESEARCHER mcp_bundle_ref hardening in policies/agent_registry.json
  • when CI maintenance changes the Python dependency audit contract or the tracked runtime report namespaces, sync the root AI/docs entrypoints in the same patch; current examples include .runtime-cache/test_output/ci/ and configs/pip_audit_ignored_advisories.json, plus the dashboard and desktop install-time ENOSPC recovery knobs plus the Docker daemon precheck retry knobs registered in configs/env.registry.json, and the bounded transient npm registry socket-timeout retries inside scripts/install_dashboard_deps.sh plus scripts/install_desktop_deps.sh; current CI credential/evidence examples also include the upstream receipt refresh fallback to scripts/verify_upstream_slices.py --mode smoke and the strict live-provider rule that resolves process env first and ~/.codex/config.toml second while keeping dotenv and shell-export fallbacks disabled on mainline; staged dashboard UI-audit workspaces must also keep package-local frontend sources inside the temporary workspace root instead of relying on out-of-root symlinks that Turbopack rejects, and repeated pnpm ERR_PNPM_ENOENT recovery should escalate to workspace-local store recovery instead of repeating the same failing fresh-store copy path
  • when retention and space-governance contracts change, sync the root AI/docs entrypoints in the same patch; current examples include retention log_lane_summary + space_bridge, serial-only heavy cleanup ordering, cleanup inventory consistency checks, and the rule that ~/.cache/codeflow is the repo-external strong-related root while shared ecosystem layers stay observe-only; current Docker runtime lane changes must keep cleanup scoped to Codeflow-owned images/containers/volumes while workstation-global Docker/cache totals stay audit-only; current machine-temp examples also include ~/.cache/codeflow/tmp/docker-ci/runner-temp-*, ~/.cache/codeflow/tmp/clean-room-machine-cache.*, and ~/.cache/codeflow/tmp/clean-room-preserve.*, which stay repo-external-related under wave3 instead of defaulting to Darwin TMPDIR; current closeout slices also include machine_cache_summary + machine_cache_auto_prune in the retention/space-governance bridge, the structured Docker runtime receipt at .runtime-cache/codeflow/reports/space_governance/docker_runtime.json, repo-owned buildx local cache under ~/.cache/codeflow/docker-buildx-cache/, plus the repo-owned singleton Chrome root under ~/.cache/codeflow/browser/chrome-user-data/ that allow_profile now attaches to over the fixed CDP endpoint instead of reusing the default Chrome root; CI / docker / clean-room lanes still fail closed back to ephemeral
  • when workflow-case / proof-pack / compare / task-pack / queue-scheduling contracts change, sync the root AI/docs entrypoints in the same patch; the current examples are .runtime-cache/codeflow/workflow-cases/, proof_pack.json, dedicated run-compare surfaces, desktop Flight Plan preview, and timezone-safe queue scheduling inputs
  • when Version B closeout work changes the public front door, shared locale substrate, read-only MCP exposure, or operator-copilot surfaces, sync README.md, docs/index.html, configs/docs_nav_registry.json, apps/orchestrator/README.md, and the root AI entrypoints in the same patch so doc-drift and doc-sync gates keep following the live Command Tower / Workflow Cases / Proof & Replay contract
  • when ecosystem-binding, builder-entrypoint, or distribution-facing surfaces change, sync the root AI/docs entrypoints in the same patch; current examples include configs/docs_nav_registry.json, the package-facing frontend-api-client / frontend-shared READMEs, and the dashboard home/docs landing sections that explain Codex / Claude Code / read-only MCP plus the first-run -> proof -> share loop
  • when a follow-up builder/adoption slice adds a legal contract-package guide or surfaces integrations/skills adoption more directly on the dashboard home, keep this file, AGENTS.md, and the root/docs entrypoints aligned in the same patch; current examples include packages/frontend-api-contract/README.md, the dashboard-home integrations section in apps/dashboard/components/DashboardHomeStorySections.tsx, and the docs-base resolver now honoring /integrations/ and /skills/
  • when a later ecosystem-adoption slice adds copy-paste starter kits or ecosystem-native example configs, keep this file, AGENTS.md, and the root/docs entrypoints aligned in the same patch; current examples include docs/agent-starters/index.html, examples/coding-agents/README.md, the shared read-only MCP example under examples/coding-agents/mcp/, and the local plugin-bundle manifests under examples/coding-agents/plugin-bundles/ that stay starter-only rather than official marketplace artifacts
  • when a later Phase 2 wave adds dedicated public sub-entrypoints (for example /ecosystem/, /builders/, /use-cases/, /compatibility/) or moves additional dashboard home hero/ecosystem/AI/builder copy into the shared locale substrate, sync the root AI/docs entrypoints in the same patch so doc-sync gates can trace the new discoverability surfaces without guessing
  • when a follow-up Phase 2 wave adds new public discoverability hubs such as /ai-surfaces/ or makes the dashboard-home locale toggle drive server-rendered copy through cookie-backed preference sync, update the root AI/docs entrypoints and release-facing docs in the same patch so doc-sync gates can follow both the public wording and the locale-contract change; current examples include docs/ai-surfaces/index.html, the extracted apps/dashboard/components/DashboardHomeStorySections.tsx narrative layer, and the AI Work Command Tower wording now shared by the dashboard metadata and the public Pages landing copy
  • when that wave also moves the dashboard-home story into a dedicated shared-copy component, keep this file, AGENTS.md, and CHANGELOG.md aligned in the same patch so quick-feedback gates can trace the locale-aware rendering path instead of inferring it from page-local literals
  • keep the root wording aligned when the dashboard home mixes cookie-backed locale SSR with client-side locale refresh, because that split is easy to miss when only the page diff is inspected
  • the current concrete examples are docs/ai-surfaces/index.html, apps/dashboard/components/DashboardHomeStorySections.tsx, packages/frontend-shared/uiLocale.ts, and dashboard metadata that now says "AI Work Command Tower for Codex, Claude Code, and MCP"
  • when the next Phase 2 wave hardens desktop Run Detail / Overview operator wording through the shared locale and shared status-presentation substrate, keep this file, AGENTS.md, and the desktop/module docs aligned in the same patch; current examples include apps/desktop/README.md, packages/frontend-shared/uiCopy.ts, and locale-aware desktop tests for RunDetailPage / OverviewPage packages/frontend-shared/uiLocale.ts, and dashboard metadata that now says "AI Work Command Tower for Codex, Claude Code, and MCP"
  • when a later Phase 2 wave hardens desktop Run Detail / Overview operator-surface locale coverage or moves more desktop strings onto @codeflow/frontend-shared, keep the root AI entrypoints aligned in the same patch; current examples include locale-aware desktop status labels, shared-copy Run Detail table/action chrome, and zh-CN regression coverage
  • when a front-door discoverability wave adds or reprioritizes public integration/skills/SEO entrypoints, keep this file, AGENTS.md, and the root/docs entrypoints aligned in the same patch; current examples include docs/integrations/index.html, docs/compatibility/index.html, docs/skills/index.html, docs/robots.txt, docs/sitemap.xml, the docs-navigation registry move that now treats ecosystem/use-cases/AI/MCP/API/builders/compatibility as primary public entrypoints, and the skills quickstart CTA shift toward in-page adoption/maintainer anchors instead of a dead public-repo tree link
  • when dashboard route-level discoverability or Workflow Case list locale coverage changes, keep this file, AGENTS.md, and the root/docs entrypoints aligned in the same patch; current examples include route metadata on apps/dashboard/app/command-tower/page.tsx, apps/dashboard/app/workflows/page.tsx, and apps/dashboard/app/workflows/[id]/page.tsx, plus the shared-copy workflow list substrate and the matching metadata/locale regression coverage in apps/dashboard/tests/command_tower_page_ssr_query_repro.test.ts, apps/dashboard/tests/workflow_detail_page.test.tsx, and apps/dashboard/tests/workflows_queue_page.test.tsx list substrate now carried through packages/frontend-shared/uiCopy.ts and packages/frontend-shared/uiCopy.js
  • when dashboard home discoverability grows a new integrations/skills adoption layer or package-contract CTA path, keep this file, AGENTS.md, and the root/docs entrypoints aligned in the same patch; current examples include apps/dashboard/components/DashboardHomeStorySections.tsx, the public-docs resolver allowlist in apps/dashboard/lib/env.ts, the matching env/home regression coverage, and the repo-owned packages/frontend-api-contract/README.md guide that now sits between the public API quickstart and the raw generated .d.ts files
  • when a later discoverability wave adds a public compatibility/adoption matrix, keep this file, AGENTS.md, and the root/docs/dashboard entrypoints aligned in the same patch; current examples include docs/compatibility/index.html, configs/docs_nav_registry.json, docs/sitemap.xml, the dashboard public-docs allowlist in apps/dashboard/lib/env.ts, and the dashboard-home integration layer now pointing teams toward a compatibility ladder before they choose protocol, skills, builders, or proof-first onboarding
  • when a follow-up discoverability wave adds public copy-paste starter kits or local bundle examples for Codex / Claude Code / OpenClaw, keep this file, AGENTS.md, and the root/docs entrypoints aligned in the same patch; current examples include docs/agent-starters/index.html, docs/examples/agent-starters/, examples/coding-agents/, configs/root_allowlist.json, and the root/docs wording that now separates host-platform plugin reality from Codeflow's own publication state
  • when a later polish wave compresses the public homepage or dashboard-home discovery stack into a clearer route page, keep this file, AGENTS.md, and the root/docs/dashboard entrypoints aligned in the same patch; current examples include the homepage mini-nav, reduced hero CTA set, compatibility-first routing, and the dashboard-home adoption layer consolidating ecosystem / integrations / AI / builders into one smaller decision surface
  • when a follow-up CTA polish slice changes that dashboard adoption layer again, keep this file, AGENTS.md, README.md, CHANGELOG.md, docs/index.html, and apps/dashboard/README.md aligned in the same patch; current examples include keeping /compatibility/ as the main routing card, restoring a lighter /use-cases/ proof-first side door, and updating the adoption-nav accessibility label so the dashboard no longer advertises the old integration-only action group
  • when the next Phase 2 wave deepens public MCP / API discoverability, keep this file, AGENTS.md, and the root/docs entrypoints aligned in the same patch; current examples include docs/mcp/index.html, docs/api/index.html, the dashboard-home AI CTA, and root navigation that points readers toward read-only MCP and API quickstarts without implying hosted/write-capable MCP shared-copy Run Detail table/action chrome, and zh-CN regression coverage
  • when Prompt 6-style skills-bundle and workflow/control-plane read-model work lands, keep this file, AGENTS.md, README.md, and the orchestrator/docs entrypoints aligned in the same patch; current examples include policies/skills_bundle_registry.json, enriched role_binding_summary.skills_bundle_ref metadata, and workflow_case_read_model on workflow/control-plane reads that remain explicitly read-only
  • when a Prompt 7-style frontend slice projects those same read models onto dashboard or desktop Workflow Case detail surfaces, keep this file, the root AI entrypoints, and the module READMEs aligned in the same patch; current examples include the read-only Workflow read model cards on apps/dashboard/app/workflows/[id]/page.tsx and apps/desktop/src/pages/WorkflowDetailPage.tsx, plus the typed frontend RoleBindingReadModel / WorkflowCaseReadModel surfaces that stay below task_contract
  • when a Prompt 8-style slice converges the OpenAPI/frontend-contract generation chain or projects role_binding_read_model onto dashboard/desktop Run Detail surfaces, keep this file, the root AI/docs entrypoints, and the module READMEs aligned in the same patch; current examples include docs/api/openapi.codeflow.json, generated @codeflow/frontend-api-contract read-model types, and the read-only Run Detail operator summaries that continue to treat task_contract as execution authority
  • when a Prompt 9-style slice turns role / bundle / runtime truth into dashboard/desktop Agents + Contracts operator catalog surfaces, keep this file, the root AI/docs entrypoints, and the module READMEs aligned in the same patch; current examples include the registry-backed /api/agents role catalog, the normalized /api/contracts inspector payload, and the same read-only authority/advisory wording carried through both web and desktop operator shells
  • when a Prompt 10-style slice turns those read-only catalog surfaces into a repo-owned role-configuration control plane, keep this file, the root AI/docs entrypoints, and the module READMEs aligned in the same patch; current examples include policies/role_config_registry.json, the role-config preview/apply routes under /api/agents/roles/{role}/config*, the generated frontend contract bindings for those routes, and the rule that Agents becomes the control desk while Contracts stays inspector-first and task_contract remains the only execution authority
  • when a Prompt 10 follow-up slice adds derived runtime capability posture to intake previews, run manifests, operator-copilot briefs, or the dashboard/desktop Contracts and Run Detail surfaces, keep this file, the root AI/docs entrypoints, and the module READMEs aligned in the same patch; current examples include execution_plan_report.runtime_capability_summary, role_binding_read_model.runtime_binding.capability, the shared dashboard/desktop runtime-capability copy, and the fail-closed wording that keeps chat-compatible lanes separate from tool execution parity
  • when a Prompt 10 closeout fix changes how contract package entrypoints load on CI/governance paths, keep this file, the root AI/docs entrypoints, and the relevant module/docs READMEs aligned in the same patch; current examples include lazy-loading codeflow_orch.contract so ContractValidator imports and schedule-boundary checks stay below runtime-provider dependencies such as httpx on Quick Feedback lanes
  • when a Prompt 10 Wave 3 slice hardens builder/client entrypoints into a repo-owned starter path, keep this file, the root AI/docs entrypoints, and the package/docs entrypoints aligned in the same patch; current examples include packages/frontend-api-client/examples/control_plane_starter.local.mjs, the package-facing createControlPlaneStarter(...) bootstrap flow, and the rule that the starter remains a repo-owned integration edge rather than a hosted SDK or marketplace claim
  • when dashboard dependency verification learns about new runtime-critical packages for quick/clean-room lanes, keep this file, AGENTS.md, and the relevant package/docs entrypoints aligned in the same patch; current examples include scripts/install_dashboard_deps.sh verifying that jsdom itself loads successfully so partial dashboard installs fail fast before the quick lane claims success
  • when the Final-100 / Wave 4 follow-up slice adds hosted pilot readiness or queue-first mutation groundwork, keep this file, AGENTS.md, and the root AI/docs entrypoints aligned in the same patch; current examples include render.yaml, configs/docs_nav_registry.json, the hosted CODEFLOW_API_ALLOWED_ORIGINS env wiring across .env.example, apps/orchestrator/.env.example, configs/env.registry.json, and configs/env_direct_read_allowlist.json, plus the rule that apps/orchestrator/src/codeflow_orch/mcp_queue_pilot_server.py and queue preview/cancel routes remain repo-owned operator groundwork instead of live hosted proof or public write-capable MCP
  • when a Final-100 hosted/operator follow-up only moves governance, queue-pilot, or API posture files after the public docs already moved, keep this file and AGENTS.md updated in the same patch so the ci-diff doc-sync gate can see the root AI navigation layer refresh above configs/env_direct_read_allowlist.json, render.yaml, and the guarded queue preview/cancel operator surfaces instead of flagging logic-only drift
  • when staged dashboard smoke builds change their dependency-install behavior or the apps/dashboard/lib/types.ts export bridge, keep this file and the root/module docs aligned in the same patch so UI-audit / pre-push failures stay attributable to staging drift vs real dashboard regressions
  • when clean-room recovery changes the package-local install order for frontend-api-client, keep this file and the root docs aligned in the same patch so recovery gates fail on product regressions instead of missing local package installs
  • when clean-room recovery changes the ordering between workspace cleanup and broad runtime deletion, keep this file and the root docs aligned in the same patch; current examples include running scripts/cleanup_workspace_modules.sh before the clean-room rm -rf sweep, plus quarantining stubborn dashboard module residue when recursive delete alone is not enough, so the recovery lane does not abort early on transient bind-mounted trees

Read the full file on GitHub · 417 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. 6d ago First seen · 417 lines · 6,065 tokens per session scan D 9d1badedb1e6

Subscribe to this mod's changes

codeflow CLAUDE.md is an instructions file published in the GitHub repository xiaojiou176-open/codeflow (207 stars, last pushed 1mo ago), licensed MIT. It adds 6,065 tokens to every session, about $0.0303 per session on Opus 5. A static security scan graded it D with 2 findings (reads agent configuration directories, recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other instructions, from other repositories

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.

vercel/next.js · 7,296 tokens

codex AGENTS.md

AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.

openai/codex · 5,182 tokens

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).

microsoft/vscode · 6,785 tokens

vscode oss-third-party-notices.instructions.md

Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).

microsoft/vscode · 5,001 tokens

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.

langchain-ai/langchain · 4,469 tokens

spec-kit AGENTS.md

AGENTS.md instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.

github/spec-kit · 7,104 tokens