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 agentmods add skills/mvschwarz/openrig/human-in-the-loopnpx skills add mvschwarz/openrig --skill human-in-the-loopgit clone --depth 1 https://github.com/mvschwarz/openrigWrote 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/mvschwarz/openrig/human-in-the-loop)<a href="https://agentmods.dev/skills/mvschwarz/openrig/human-in-the-loop"><img src="https://agentmods.dev/badge/skills/mvschwarz/openrig/human-in-the-loop.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 | $0.00092 | $0.01156 |
| Opus 5 | $0.00046 | $0.00578 |
| Sonnet 5 | $0.00018 | $0.00231 |
| Haiku 4.5 | $0.00009 | $0.00116 |
Grade A, and why
human-in-the-loop 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 5d 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 — 105 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Human In The Loop
The primitive that treats humans as durable network participants — attention surfaces, queues, decision records, routing semantics — not as ad-hoc chat receivers.
Autonomy is not the absence of humans; it is knowing when human judgment is needed and making that handoff crisp.
Use this when
- A slice closeout needs classifying: auto-continue, human gate, or park
- A real decision needs to land in front of a human (usage limits, provider auth, roadmap tradeoff, product-intent ambiguity)
- Designing a human queue/dashboard surface
- Returning a hot potato to orchestration after human approval
Don't use this when
- The slice closeout is clean and
PROGRESS.mdalready names the next safe slice. Default RSI conveyor continues; do NOT manufacture a human gate. - The escalation is just a status update. Humans are participants for decisions, not narration.
- The next owner is another agent. Use queue-handoff, not human-in-the-loop.
The 3-class closeout classification
In a productized daemon-backed version, closeout classifies the next step BEFORE touching the human queue:
| Class | When | Action |
|---|---|---|
| auto-continue | Slice closes cleanly, next named slice in workstream plan | Mark closed; create next-owner qitem from plan |
| human gate | Genuine decision needed (usage limits, provider auth, product-intent ambiguity, roadmap tradeoff) | Create human queue item with proof + decision text + recommended default + action outcomes |
| park | Intentionally stop the conveyor (e.g., waiting on external) | Stop with reason + resumption path |
Failure modes (5)
- Human decision needed, but the rig only mentions it in chat. Decisions belong as durable attention items, not chat messages.
- Human queue item lacks enough plain-English context for a decision. Include proof + decision text + recommended default + action outcomes.
- Human response updates a file but does not wake the next owner. Approval should return the hot potato; feedback should create the next durable qitem.
- The dashboard shows too much raw rig state and hides the actual decision queue. Decision queue is the primary surface; rig state is secondary.
- A clean closeout is parked on the human even though
PROGRESS.mdalready names the next safe slice. Don't manufacture human gates.
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.
- 5d ago First seen · 105 lines · 92 tokens per session scan A 14cf0c4a0dc8
human-in-the-loop is a skill published in the GitHub repository mvschwarz/openrig (65 stars, last pushed today), licensed Apache-2.0. It adds 92 tokens to every session and 1,156 once invoked, about $0.0005 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 2026-08-30.
Other skills, from other repositories
autoprompt
Explicit-only useful-first orchestration. Invoke only when the user names autoprompt - typed as /autoprompt or in plain language such as "act in autoprompt mode" - to turn a mission into one executable roadmap, build dependency-safe lanes, and verify the result with independent reviewers. Do not infer invocation from…
autoprompt
Explicit-only useful-first orchestration. Invoke /autoprompt to turn a mission into one executable roadmap, build dependency-safe lanes, and verify the result with independent reviewers. Never infer invocation from ordinary requests. Never resume from leftover artifacts without an explicit resume instruction.
autoprompt
Explicit-only useful-first orchestration. Invoke /autoprompt to turn a mission into one executable roadmap, build dependency-safe lanes, and verify the result with independent reviewers. Never infer invocation from ordinary requests. Never resume from leftover artifacts without an explicit resume instruction.
ap-implementer
L3 executor - G4 IMPLEMENT. Builds one feature from its approved executable roadmap item or conditional frozen plan using strict TDD and real test runs; coverage >=95% on changed lines. Reports PLAN-CONFLICT rather than improvising.
ap-execharness-resolver
L3 executor - EXECHARNESS RESOLVE. Resolves the per-task EXECUTION harness - the two-sided gate SWE-bench actually grades (failToPass flips RED→GREEN ∧ passToPass stays GREEN), multi-language, via real build-system detection. Ingests shipped FAILTOPASS/PASSTOPASS, else derives failToPass from the mission's behavioral…
ap-feature-coordinator
L1 feature coordinator - drives approved ROADMAP.md lanes through their required build/review/verification gates and owns the run-wide feature frontier.