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 jongwony/epistemic-protocols --skill onboardgit clone --depth 1 https://github.com/jongwony/epistemic-protocolsWrote 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/jongwony/epistemic-protocols/onboard)<a href="https://agentmods.dev/skills/jongwony/epistemic-protocols/onboard"><img src="https://agentmods.dev/badge/skills/jongwony/epistemic-protocols/onboard/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/jongwony/epistemic-protocols/onboard"><img src="https://agentmods.dev/badge/skills/jongwony/epistemic-protocols/onboard.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Data Exfiltration · line 419 Code or instructions that leak agent conversation context to external services, potentially exposing sensitive user interactions.Fix: Remove any code that sends prompts, responses, or session data externally. Preserve user privacy; never exfiltrate conversation content.
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 | $0.00022 | $0.06792 |
| Opus 5.5 | $0.00009 | $0.02717 |
| Sonnet 5.5 | $0.00004 | $0.01358 |
| Haiku 4.5 | $0.00002 | $0.00679 |
Grade A, and why
onboard 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 — 425 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Onboard Skill
Start with a quick recommendation based on recent sessions, then optionally continue to guided learning — so users experience value first, learn second.
Invoke directly with /onboard when the user wants onboarding or protocol discovery.
When to Use
Invoke this skill when:
- A new user wants to discover which epistemic protocols fit their workflow
- A user wants to experience protocols through guided practice
- Re-onboarding after new protocols are added or workflow changes
Skip when:
- User already knows which protocol to use (direct invocation)
- Quick single-protocol question (answer directly)
Workflow Overview
Quick Proof: ENTRY → QUICKSCAN → PICK → EVIDENCE → TRIAL → INSIGHT → NEXT
Targeted: ENTRY → QUICKSCAN → MAP → SCENARIO → TRIAL → QUIZ → GUIDE
Targeted + std: ENTRY → SCENARIO → TRIAL → QUIZ → GUIDE
| Phase | Owner | Tool | Purpose |
|---|---|---|---|
| 0. Entry | Main | Gate | Path selection: quick/targeted |
| 1. Quick Scan | Main | Glob, bounded head read | User Context Profile extraction |
| 2a. Pick | Main | — | Quick path: recommendation |
| 2b. Evidence | Main | — | Quick path: evidence card |
| 2. Map | Main | — | Targeted path: Profile → Protocol matching |
| 3. Scenario | Main | Gate | Targeted path: context-personalized intervention point |
| 4. Trial | Main | Gate | Real protocol execution (quick: mini trial, targeted: full trial) |
| 4→Q. Insight | Main | — | Quick path: post-trial insight card |
| 4→Q. Next | Main | Gate | Quick path: simplified navigation |
| 4→5 LOOP | Main | Gate | Targeted path: post-trial navigation |
| 5. Quiz | Main | Gate | Targeted path: Socratic protocol recognition quiz |
| 6. Guide | Main | Gate | Targeted path: summary + next protocol suggestion |
Data Sources
Compact mapping for inline use.
| Protocol | When to Use | Key Patterns |
|---|---|---|
Aitesis /inquire |
A task rests on missing context or unchecked assumptions — collect what the AI can reach on its own, hand back what it cannot as the user's own unknown | Hand-off or execution about to start on context nobody gathered; implicit requirements, environment dependencies, prior decisions (for prior-session recall → use /recollect) |
Euporia /elicit |
Intent articulated but the decisions it turns on not yet named | Multi-axis intent without single axis-specific protocol fit; coordinates surface from the user's material, words, and the domain's usual decisions |
Heuresis /ideate |
Object-level candidate field is empty or has prematurely converged — widen it before any selection is made | Zero entry questions (blank, seeded, or resumed, inferred without asking — for a bare /ideate, from the conversation before it), frame-first while the field holds no candidate, no elimination or ranking during generation, every candidate tagged with where it came from (the user, the AI, or the tag named material came with, else the turn it was read from) |
Proplasma /preview |
Right before a direction commitment when the candidates cannot be judged from descriptions — contrast cheap discard-committed placeholder probes on AI-drafted axes relayed with their basis | Principle-delegation at direction gates ("go with the recommended direction"), option-set reconstruction instead of choosing, "I'd have to see it" decision stalls |
Hypotyposis /sketch |
A form has to be made, intent cannot yet be settled from descriptions, and the user would recognize it on sight — relay what the next versions try, show each version before the standing, revise from the versions the user's words point to, end when they recognize one they met | "I'd know it when I see it", a plan stalled at its first draft, a description rewritten repeatedly instead of made |
Analogia /ground |
Auditing what a mapping being relied on licenses about a case, reading any account it can reach as evidence | Intended conclusions whose structural evidence or limits remain uncertain |
Periagoge /induce |
Concrete cases sharing something not yet named, or a name being reconsidered from its cases — crystallize the abstraction they share | 2+ concrete cases sharing something whose name, rule, or reach is unsettled, a named abstraction being reconsidered from its cases included — the question is what abstraction the given cases share, where /ground asks what carrying a given structure to a target supports, or tests an abstraction against its own cases |
Merismos /apportion |
An autonomous goal is stated but its unit plan is uncompiled — cut it into coarse units at cited seams and close each unit before the run begins — a derived completion condition where one compiles, a recorded acceptance or a recorded reservation where none does — then hand the taken plan to the run | An unattended-run directive ("work through", "go through all", "run until done"), a goal that plainly exceeds one execution horizon, or a stop-hook being configured — each only shows an autonomous interval is intended, so the deciding check is whether the goal already carries units whose completion conditions are settled — closed by a determinate predicate, by a recorded acceptance, or by a recorded reservation; where one already does, /apportion shows that plan on its first sheet with that reading as a contrary ground |
Epharmoge /contextualize |
A result — this session's or another's — is correct but may not fit where it lands or may leave out an intent the conversation stated | Misfits and omissions against everywhere the result lands and the session-built constraints, shown on one sheet |
Horismos /bound |
Deciding what to delegate to AI | Provisional whole map, progressive examination, source-bound settlement and residual |
Anamnesis /recollect |
Resolving vague recall of prior sessions or discussions — one session, or a line of work, topic, or settled concept spread across several | Cross-session state recovery via narrative recognition (Recognition over Retrieval); a unit above one session is composed from its deposits at read time |
Katalepsis /grasp |
Something in play you can't follow yet, or nod along to unsure — a map of understanding over your purpose, each aspect saying what it rests on | User-intent grasp for a target present in context and quotable, whoever produced it — reviews, plans, papers, docs, or code changes |
Hyphegesis /conduct |
Conducting the method of a multi-move work prospect before object-level cognition — order, independence, combination, stopping, where results go | Multi-move work with non-trivial conduct (migrations, staged investigations, entangled adversarial/parallel/synthesis); "how should I approach this whole thing?" meta-questions |
What ships with it
2 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.
- 7d ago Changed · -3 lines f1c621116a43
- 13d ago Changed de511d0c768a
- 14d ago Changed 699b948825f7
- 16d ago Changed · -3 lines f2d8a32f534d
- 17d ago Changed ae99d1f52565
- 19d ago Changed · -1 lines dabef63ae1fc
- 20d ago Changed 6d3e6fe42cd3
- 24d ago Changed cb282cf1d69b
- 25d ago Changed · -2 lines 5dd082e19759
- 27d ago First seen · 434 lines · 22 tokens per session scan A 93899b9605d9
onboard is a skill published in the GitHub repository jongwony/epistemic-protocols (174 stars, last pushed yesterday), licensed MIT. It adds 22 tokens to every session and 6,792 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 2026-08-30.
Other skills, from other repositories
rational-buddhism
A way to examine spiritual, supernatural, or other difficult-to-test claims using skepticism, personal testing, and scientific standards.
flow-next-audit
Audit .flow/memory/ entries against current code and keep, update, consolidate, replace, delete, or harden each. Use when asked to audit memory or graduate a recurring lesson into a gate.
maestro-coach
Decision support and grounded teaching for the human. Invoke when the user signals a message did not land ("wait, what?", "I don't understand", "explain it like I'm five", "which is better?", "I'm not sure") about a question or fork they have been asked, or when they hand over a locked decision, a lesson, or a bundle…
no-vibe-challenge
One coding challenge built from the current syllabus or the project. The user writes all the code. The agent gives the task, hints when asked, and reviews when asked.
no-vibe
Learning mode. Use ONLY when .no-vibe/active exists at the project root or the user invoked /no-vibe. The agent teaches in chat and never writes project files. Every reply starts with the header [no-vibe] Step: · Section: · Exercise: .
research-mentor-review
Provide constructive, evidence-based research mentoring feedback on student manuscripts, proposals, analyses, code, figures, and presentations. Use for advisor-style review that teaches the next improvement without impersonating an advisor or making unsupported scientific judgments.