friction-audit
Skill1 starsChanged 5 months agoMITView on GitHub
Audit a product flow for friction — every point where the system makes the user do work it could do itself — then design the automatic path for each. Classifies friction (ask-for-permission, re-do/re-send, manual recovery, re-discovery, dead-end, redundant confirmation), decides what should become automatic with a safety carve-out for destructive/irreversible/cross-user actions, and names the tradeoff. Triggers on "audit this flow", "where's the friction", "why does the user have to", "make this automatic", "remove the friction", "this has too much friction", "it should just happen", "reduce steps", "the user shouldn't need to". Use when reviewing a UX flow, agent loop, or feature for needless user effort — not for visual/styling review (use ui-ux for that).
In this repository51 other files it gives its agents.
- architect-enterprise-aiSkill
- author-agent-skillsSkill
- build-agent-harnessesSkill
- build-aiewf-2024-ragSkill
- build-generative-mediaSkill
- build-multimodal-productsSkill
- build-rag-search-stacksSkill
- build-robotics-aiSkill
- build-with-deepmindSkill
- build-with-microsoft-aiSkill
- build-with-openaiSkill
- deploy-public-sector-aiSkill
- design-agent-skillsSkill
- evaluate-with-braintrustSkill
- lead-ai-engineering-teamsSkill
- lead-ai-transformationSkill
- learn-summit-2023-remoteSkill
- learn-summit-2023-talksSkill
- navigate-aie-europe-2026Skill
- navigate-aiewf-2024-keynotesSkill
- observe-ai-productionSkill
- operate-agent-memorySkill
- operate-legal-ai-agentsSkill
- operate-openclaw-agentsSkill
- optimize-llm-inferenceSkill
- plan-sovereign-aiSkill
- price-ai-productsSkill
- run-aiewf-2024-evalsSkill
- run-llm-evalsSkill
- run-summit-2023-workshopsSkill
- secure-ai-agentsSkill
- ship-ai-as-pmSkill
- ship-aiewf-2024-agentsSkill
- train-llms-locallySkill
- architecture-reviewSkill
- chat-failure-auditSkill
- create-a-skillSkill
- cross-model-reviewSkill
- cto-weekly-reviewSkill
- ig-carouselSkill
- pdf-reportSkill
- raise-prSkill
- reel-editSkill
- research-reportSkill
- ship-checkSkill
- spec-grounded-designSkill
- state-lifetime-decisionSkill
- ste100Skill
- scroll-engineeringSkill
- ui-uxSkill
- verify-like-a-userSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 3 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 52
Read the actual code or transcript — do not infer the flow from memory or docs.
- Distinctive: line 59
Read friction-taxonomy.md.
- Very specific: line 68
Not every user action is friction.
- Very specific: line 101
Write the report from audit-template.md to the output dir and update index.md.
- Distinctive: line 107
Never edit code or commit here — this skill audits and designs.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Friction Audit
Find every place a flow asks the user to do something the system could do itself, then design the automatic path. The bias: anything the system can do for the user, it should — unless it's destructive, irreversible, or crosses a trust boundary.
Works with any agent. No vendor APIs. Reads the code/flow you point it at.
Supporting files (read when needed):
- friction-taxonomy.md — the six friction classes, the auto-vs-manual rule, the tradeoff prompts
- audit-template.md — the inventory + auto-path report format
Output: {SKILL_OUTPUT_DIR}/friction-audit/ — see ../OUTPUT.md
Step 0 — Scope the flow
Establish what you're auditing (infer from the request; ask only if missing):
- The flow — a feature, an agent loop, an onboarding path, a described journey, or a
real transcript (pair with
chat-failure-auditif you only have symptoms, not a flow). - The actors — what is the user, what is the system/agent, what is external (a third-party API the system can't control).
- Authority already granted — what has the user already connected / consented to / configured? Friction often = re-asking for authority the user already gave.
Resolve output dir per ../OUTPUT.md. Default ./skill-outputs/friction-audit/.
Step 1 — Walk the flow, mark every user action
Trace the flow step by step. At each step where the user must act, record it as a candidate. For each, capture: what the user does, why the system requires it today, and what boundary the action sits behind (none / authority / destructive / external).
Read the actual code or transcript — do not infer the flow from memory or docs. If it's code, follow the real control path (the handler, the prompt, the state writes).
Step 2 — Classify each friction point
Read friction-taxonomy.md. Tag each candidate with one class:
- Ask-for-permission — system asks to do something the user already authorized.
- Re-do / re-send — user repeats an action because state didn't cross a boundary.
- Manual recovery — user reconnects/refreshes/retries what the system could self-heal.
- Re-discovery — user or model re-finds something already found.
- Dead-end — system fails or stops without offering the next step.
- Redundant confirmation — confirmation gating a non-destructive, reversible action.
Not every user action is friction. A deliberate choice, a destructive confirm, or a real
external dependency is legitimate — mark those keep with the reason.
Step 3 — Decide auto-vs-manual (the safety carve-out)
For each friction point, apply the rule from the taxonomy:
Make it automatic unless the action is destructive, irreversible, crosses a user/ tenant boundary, spends real money/quota without a cap, or the user genuinely wants the choice. Those stay manual — with a confirm or an undo window, not silent.
State the verdict (automate / keep / automate-with-guardrail) and the reason.
Step 4 — Design the automatic path
For each automate verdict, design the fix in one short block:
- Trigger — what now causes the system to act (an event, a step boundary, a turn).
- Mechanism — where the work moves (often: carry state across the boundary the user was bridging by hand; self-heal instead of surfacing the error; do it in-loop).
- State — if it introduces or changes persisted state, hand off to
state-lifetime-decision(scope × durability × isolation) rather than guessing. - Tradeoff — name the cost in one line (latency, cache, context bloat, surprise) and why it's worth it. No fix is free; say what you're paying.
Step 5 — Output to user
Write the report from audit-template.md to the output dir and update
index.md. Then in chat:
- The friction inventory as a table (point · class · verdict).
- The top 1–3 auto-paths to ship first, ranked by user pain × ease.
- Anything you marked
keepand why (so the user can challenge it). - Never edit code or commit here — this skill audits and designs; implementation is a separate, confirmed step.
Edge cases
- Only a transcript, no flow — run
chat-failure-auditfirst to extract failures, then audit the flow those failures live in. - Everything looks like friction — you're probably tagging legitimate choices as
friction; re-check Step 2's
keeprule. A flow with zero user input is usually wrong. - The fix needs new state — don't inline a scope/TTL guess; defer to
state-lifetime-decision. - Big fork in the fix (build vs buy, two mechanisms) — defer to
architecture-review. - Destructive action with no confirm today — that's reverse friction (too little); flag it, recommend adding a confirm/undo.
Invocation examples
@friction-audit our connector-enable flow — why does the user have to re-send?
audit this onboarding path for friction
where's the friction in the agent's tool-calling loop?
this should just happen automatically — find every place it doesn't
remove the friction in document sharing, but keep deletes safeDoes it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

