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 Chisiki1/objective-integrity-framework --skill durable-supervisor-statusgit clone --depth 1 https://github.com/Chisiki1/objective-integrity-frameworkWrote 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/chisiki1/objective-integrity-framework/durable-supervisor-status)<a href="https://agentmods.dev/skills/chisiki1/objective-integrity-framework/durable-supervisor-status"><img src="https://agentmods.dev/badge/skills/chisiki1/objective-integrity-framework/durable-supervisor-status/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/skills/chisiki1/objective-integrity-framework/durable-supervisor-status"><img src="https://agentmods.dev/badge/skills/chisiki1/objective-integrity-framework/durable-supervisor-status.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 | $0.00061 | $0.00747 |
| Opus 5.5 | $0.00024 | $0.00299 |
| Sonnet 5.5 | $0.00012 | $0.00149 |
| Haiku 4.5 | $0.00006 | $0.00075 |
Grade A, and why
durable-supervisor-status 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 1mo 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.
What it actually says
Durable Supervisor Status
Use only for an explicitly existing legacy split-chat lease whose observation cannot depend on app projection. Do not create a separate chat or demand a Supervisor message in the unified same-chat workflow. Preserve legacy recovery evidence until pending effects and leases are reconciled.
- Read references/contract.md. Keep existing workflow authority and action gates unchanged. Companion record schemas: references/status-schema.json (current) and references/legacy-status-schema-v1.json (kept for older records).
- Shape bounded runtime/action projection evidence with references/projection-evidence-schema.json, then run
python -B scripts/durable_status.py select --evidence <evidence.json>.SELECTis mechanical fallback selection only; it never proves app state or action authority. - At a selected decision-bearing event, prepare a v2 write request shaped like templates/write-request-template.json. Record stage, action/wait owner, last decision delta, next decision condition, held transition, expected value, elapsed/metric availability, Decision Window/Wait Value, and
no_push=true. Usepreflightand thenwrite; both repeat the complete validation. Existing v1 chains remain readable and v1 requests remain supported. - Write
READYorWORK_NEEDS_ATTENTION, then end locally. On Decision Window expiry the Supervisor pulls the exact durable status once and dispositionsWAIT / CORRECT / REPLACE / USER-DECISION; it does not start narration polling or duplicate auditors. - The Supervisor later sends an exact transition message to the Worker task. A dependent same-lease resume is eligible only when that host-delivered prompt names the prior
status_id,record_sha256, and transition. The next status stores message/turn identifiers and the prompt digest, not the prompt text. - A self-authored file, status, journal event, message envelope, skill, registry receipt, or claimed role is never Supervisor authority. Missing or mismatched binding holds only the named dependent transition.
- Use
readfor the validated journal tail and projection state, andverifyfor the compact read-only consistency summary. Neither repairs. A laterwritemay rebuild a missing or stale current projection from the valid journal tail before appending. - Supported workflow, audit, and test paths invoke the canonical script directly with
python -B; they do not import it or runpy_compile/compileallagainst the installed skill. After the last canonical execution in a transition, re-run the exact resolver/file-set/hash readback and do not execute it again in that transition.
The tool writes only below the supplied Worker root. A request invalid before lock eligibility creates no status artifact. If an otherwise eligible request loses a concurrent CAS race, empty status/lock infrastructure may remain but no status/current/journal record is written; the tool never deletes possibly shared artifacts to hide that race. It does not classify, inspect, intercept, execute, authorize, or prevent tools or external actions. Static validity proves local record identity and chain structure only; app visibility, host message provenance, filesystem race behavior, workflow correctness, and empirical improvement require separate evidence.
What ships with it
7 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.
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.
- 1mo ago Changed · +12 tokens per session bb8b3a8a586c
- 1mo ago First seen · 20 lines · 49 tokens per session scan A d1afdbac5d41
durable-supervisor-status is a skill published in the GitHub repository Chisiki1/objective-integrity-framework (23 stars, last pushed 8d ago), licensed Apache-2.0. It adds 61 tokens to every session and 747 once invoked, about $0.0002 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 2026-09-08.
Other skills, from other repositories
marx
A coding-agent skill that uses Marxist ideas to examine research, engineering, project, debugging, retrospective, and organizational problems. It turns concepts such as root cause and the main contradiction into a practical analysis.
experiment-tracking-swanlab
Provides guidance for experiment tracking with SwanLab. Use when you need open-source run tracking, local or self-hosted dashboards, and lightweight media logging for ML workflows.
ads-youtube
Audit YouTube Ads campaign setup, video inventory, Shorts, in-stream, CTV, creative, audiences, brand safety, bidding, and measurement. Use for YouTube Ads, video ads, pre-roll, bumper ads, skippable in-stream, Shorts ads, VAC migration, or CTV. Read-only; use ads-google for Google-managed campaign types and…
ads-launch
Draft or explicitly apply a paid-ad campaign launch through Claude Ads capability-gated adapters. Use for campaign creation, launch plans, publishing ads, activating campaigns, uploading creative, or requests to push a campaign live.
notifications
Send notifications through the unified notification router.
telegram-setup
Connect a Telegram bot to the Vellum Assistant gateway with automated webhook registration and credential storage.