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.
npx skills add GxFn/Wakeflow --skill wakeflow-governancegit clone --depth 1 https://github.com/GxFn/WakeflowWrote 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/skills/gxfn/wakeflow/wakeflow-governance)<a href="https://agentmods.dev/skills/gxfn/wakeflow/wakeflow-governance"><img src="https://agentmods.dev/badge/skills/gxfn/wakeflow/wakeflow-governance.svg" alt="Measured on agentmods" 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 | $0.00084 | $0.03667 |
| Opus 5 | $0.00042 | $0.01834 |
| Sonnet 5 | $0.00017 | $0.00733 |
| Haiku 4.5 | $0.00008 | $0.00367 |
Grade A, and why
wakeflow-governance 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 7d 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.
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 — 193 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Wakeflow Governance
This skill holds detailed Wakeflow procedures that are too bulky to keep fully resident in CLAUDE.md.
Scope
For workspace initialization or setup requests, use the Wakeflow MCP capability tool first:
- Call
wakeflow_maintain_workspacewithaction: "fresh-initialize",mode: "preview", and one closedrequest.selectionfor a fresh workspace. The selection explicitly partitionsprogram,topology,storage,governance, andhosts; repository, support-surface, and window entries use request-localselectionKeylinks. Wakeflow allocates the durable typed IDs. - Preview is read-only. Review its blockers, exact
confirmedActionPlan, returnedconfirmedActionPlanDigest, andlaunchIntents. Apply only after the user confirms the write boundary, by sending the same root and action withmode: "apply", the exact confirmed plan, and that returned digest. - Use
action: "reconfigure"only for an intentional desired-model change andaction: "reconcile"only to restore managed bytes/projections from current v3 authority. Both follow the same preview-before-apply rule.recoverrequires the exact plan, digest, and incomplete mutationoperationId. - Do not infer Design/Test from similar existing directory names such as
<WorkspaceName>Design,<ProductName>Design,<WorkspaceName>Test, or<ProductName>Test. Unless the user explicitly names those as Design/Test, Wakeflow should create/use freshDesignandTestsupport surfaces. - If a workspace contains legacy config/runtime/ledger/tool surfaces or unclear ownership, do not force fresh initialization or pass discovery/reset aliases. Report the stable blocker and use the explicit unregistered migration path only when the user has requested migration.
- During apply, Wakeflow synchronizes the workspace
.gitignoreso only.wakeflow-active/and.wakeflow-local/are ignored runtime directories. Do not add product repositories, Design/Test, ledgers,.DS_Store, or other user workspace noise as Wakeflow-generated gitignore entries. - Window removal is a reconfigure decision expressed in the complete desired v3
topology; there is no
excludeWindowsor semantic window-name mutation alias. - Pass
language: "zh"when the user is working in Chinese,language: "en"when the user asks for English, and leavelanguage: "auto"only when there is no clear preference. The returneddisplayTitleis the canonical window title: the tmux window NAME is the displayTitle (for example "AppRepo Work"), and it may appear in the tmux status bar and terminal tabs. Only the v3 activation owner may apply or restore that display metadata; title text is never identity authority. - After the confirmed apply succeeds, retain the matching preview's
launchIntents. They are host-neutral authorizations. Route each exact launch through the packaged v3 Claude host facade'slaunch-windowowner, then callwakeflow_register_window operation=registerwith the final real session handle. Registration writes one typed host-local binding and refreshes the redacted window-runtime projection; it never classifies a startup reply as ready/pending/failed. Retired public-v2 registry/window-host commands are not aliases. If the exact host effect or receipt is unavailable, report the launch intents and stop instead of writing substitute runtime files. - To rebuild selected windows, call
wakeflow_replace_windowswithoperation: "replace"and its exact typed request. Execute only the returned host-neutral replacement intent through the v3 host owner, then register the final session handle. Do not rewrite unrelated bindings or store role, cwd, or title as identity authority. - Do not use
wakeflow_maintain_workspacefresh initialization as a refresh path for window context bloat. Replacement returns only the authorized host-neutral intent; launch only that window and register its final session handle. - Pod creation and recovery are separate. Use
wakeflow_pod_open operation=launch-preview/launch-applyonly for first materialization andoperation=inspect-materializationfor an existing binding. Missing or ambiguous identity stops; never fall back to mainline or a discovered same-named worktree. Claude Code desktop windows are not an automation transport. - tmux windows cannot answer permission prompts while the user is away.
Per-repository
.claude/settings.jsonallowlists, or an explicit--claude-arg --permission-mode=acceptEditsat launch, are the user's decision; Wakeflow never chooses silently. - Do not replace that tool with a hand-written inspection checklist when the MCP server is available.
- If Wakeflow MCP tools are unavailable, say that the MCP server is unavailable and stop for plugin reload/reinstall instead of pretending initialization can proceed through docs alone.
- Wakeflow MCP initialization does not place real session handles in tracked
docs or prompts. The typed binding is the host-local identity authority;
window-runtimeis a redacted projection, not a second authority.
What ships with it
13 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- references/agents-rule-map.md 2.5 KB
- references/design-test-skill-realization-source-map.md 29 KB
- references/direct-thread-window-config.md 4.9 KB
- references/phased-migration.md 1.3 KB
- references/script-pipeline.md 9.9 KB
- references/skill-writing-style.md 3.7 KB
- references/stage-route-map.md 13 KB
- references/testing-validation.md 4.6 KB
- references/todo-backlog.md 2.1 KB
- references/wakeflow-architecture.md 4.7 KB
- references/wakeflow-delivery.md 6.5 KB
- references/wakeflow-ledgers.md 6.5 KB
- references/window-dispatch.md 11 KB
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.
- 7d ago First seen · 193 lines · 84 tokens per session scan A 3e2e739492e6
wakeflow-governance is a skill published in the GitHub repository GxFn/Wakeflow (5 stars, last pushed 4d ago), licensed MIT. It adds 84 tokens to every session and 3,667 once invoked, about $0.0004 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.
Other skills, from other repositories
fleet
Parallel campaign orchestrator. Runs multiple campaigns in coordinated waves within a single session. Spawns 2-3 agents per wave in isolated worktrees, collects discoveries, shares context between waves. Use when work decomposes into 3+ independent streams that can run simultaneously.
archon
Autonomous multi-session campaign agent. Decomposes large work into phases, delegates to sub-agents, reviews output, and maintains campaign state across context windows. Use for work that spans multiple sessions and needs persistent state, quality judgment, and strategic decomposition.
daemon
Continuous autonomous operation mode. Keeps campaigns running 24/7 by chaining Claude Code sessions via RemoteTrigger. Each session picks up from the campaign's continuation state, works until context runs low or the phase completes, then schedules the next session. Auto-stops on campaign completion or budget…
dashboard
Real-time harness observability dashboard. Reads campaigns, fleet sessions, telemetry, and pending queues to present a snapshot of harness state at a glance. Invoked by /dashboard, /do status, or phrases like "what's happening" and "show activity".
evolve
Research-driven multi-cycle improvement director. Forms causal hypotheses about why scores are low, validates them with scout agents before attacking, dispatches axis-parallel fleet attacks, extracts transferable patterns, and runs indefinitely within a budget envelope. Accumulates a persistent belief model and…
improve
Autonomous quality improvement loop. Scores a target against a rubric, selects the highest-leverage axis, attacks it, verifies, documents, and loops. No pre-planning between iterations — each loop re-scores from scratch.