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 Tom-R-Main/openFunctions --skill durable-workgit clone --depth 1 https://github.com/Tom-R-Main/openFunctionsWrote 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/tom-r-main/openfunctions/durable-work)<a href="https://agentmods.dev/skills/tom-r-main/openfunctions/durable-work"><img src="https://agentmods.dev/badge/skills/tom-r-main/openfunctions/durable-work/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/tom-r-main/openfunctions/durable-work"><img src="https://agentmods.dev/badge/skills/tom-r-main/openfunctions/durable-work.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 | $0.00029 | $0.00630 |
| Opus 5.5 | $0.00012 | $0.00252 |
| Sonnet 5.5 | $0.00006 | $0.00126 |
| Haiku 4.5 | $0.00003 | $0.00063 |
Grade A, and why
durable-work 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 2d 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
Use this workflow only when the user explicitly selects a work item covered by the configured local work profile. Installing this portable plugin does not install npm dependencies or grant execution authority. Keep it disabled until the user configures the runtime and enables it.
Before execution, use of_work_prepare to read the work contract and verification plan. Present the concrete scope and required evidence. A trusted user issues execution and completion grants with the compiled OpenFunction CLI against that exact preview; never fabricate authority from conversation text or a model-supplied approval field. Hermes approval permits a host call, while OpenFunction still checks persisted grants.
The trusted user's terminal workflow is:
openfunction work prepare --config <absolute-config-path> --work-item <selected-id> --out <absolute-private-preview-path>
openfunction work authorize --config <absolute-config-path> --preview <absolute-private-preview-path>
The authorization command displays the bound contract, scope, steps, proof methods, expiry, and limits. The user types AUTHORIZE for execution and separately types COMPLETE for automatic completion. Any other execution answer refuses; automatic completion stays off by default. This CLI is a trusted issuance path and must not become a model-callable approval tool.
Inspect and revoke persisted grants using openfunction work grants list --config <absolute-config-path> and openfunction work grants revoke --config <absolute-config-path> --grant <grant-reference>. Revocation takes effect at the next admission and completion check; it does not reverse effects already performed.
Use of_work_start, then of_work_step for the configured actions. Use of_work_verify to collect actual evidence. of_work_status inspects the attempt without silently executing or renewing it. Use of_work_finish only after the required passing evidence covers the current subject and completion authority is valid. Changed subject state requires fresh verification.
After an interruption, inspect status and use of_work_reconcile. A saved known result is replayed without repeating its effect. An uncertain effect requires reconciliation and prevents automatic completion. Do not retry unknown actions or claim completion from model narration. Revoked or expired grants, lease loss, missing proof, changed scope, or a review requirement stop further admissions.
Only actions routed through the OpenFunction executor and controller receive its receipts and recovery guarantees. Native Hermes terminal, browser, and file actions remain outside that journal. Report this coverage accurately. Keep toolsets and instructions stable during a conversation; refresh work state using tool results.
The MCP server reads ${PLUGIN_DATA}/config.json in caller-owned profile state. Credentials belong to the host's profile-scoped secret environment, never this manifest or chat output. The runtime must be installed and version-tested before plugin enablement; do not use an execution-time npx download.
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.
- 2d ago First seen · 29 lines · 29 tokens per session scan A 732315ca0bc6
durable-work is a skill published in the GitHub repository Tom-R-Main/openFunctions (4 stars, last pushed 2d ago), licensed MIT. It adds 29 tokens to every session and 630 once invoked, about $0.0001 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 2026-10-14.
Other skills, from other repositories
interview-question-preparer
Prepare an evidence-backed blog/media interview preparation brief for a specified person by researching public context, reflecting on information sufficiency, and writing the final Markdown deliverable.
Market Research Brief
Generate a structured market research brief from a feature idea and supplied business context: market landscape, competitor intelligence, opportunity assessment, and a go-to-market recommendation. Use for market, research, competitor, and product launch requests.
Smart Code Review
Review a PR diff: triage its severity, then produce a depth-appropriate review with structured findings, fix suggestions, and a merge decision. Use for code review, review PR, and severity triage requests.
Deep Research
Produce a deep, structured research report on a topic: decompose into key dimensions, analyze each with evidence and reasoning, synthesize cross-cutting insights, and surface open questions. Use for deep research, analysis, and literature/landscape reviews.
Product Launch Press Kit
Generate a product launch press kit from a product brief: a market positioning brief and a launch risk register with mitigations. Use for press kit, launch, positioning, and press release requests.
Research Report Evaluator
Evaluate a research report against its stated topic across six dimensions (relevance, completeness, source authority, depth balance, consistency, decision quality), assigning a conceptual quality level and issues to each. Use for report evaluation, research review, and quality assessment.