agentleFS
45ck45ck/hci-review-skill/.claude/skills/failure-path-audit/SKILL.md

failure-path-audit

Skill19 starsChanged 6 days agoMITView on GitHub

Audit all failure paths, edge cases, and non-happy-path states -- empty states, validation errors, permission failures, loading, undo, and recovery.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 5 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use ultrathink. Happy paths make prototypes look coherent.

    line 11
  2. Very specific:

    Write docs/hci/failure-path-audit.md unless the user asked for another path.

    line 14
  3. Distinctive:

    Create the directory if it does not exist.

    line 15
  4. Distinctive:

    Do errors appear on submit or on blur?

    line 47
  5. Distinctive:

    Delete: is it soft delete or permanent?

    line 84
  6. Distinctive:

    Every happy path implies at least 3-5 failure paths.

    line 111

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Audit failure paths and edge cases for this scope:

$ARGUMENTS

Use ultrathink. Happy paths make prototypes look coherent. Failure paths reveal whether they actually are. Review the actual codebase, components, error handling, and UI states.

Output

Write docs/hci/failure-path-audit.md unless the user asked for another path. Create the directory if it does not exist.

Required categories

1) Empty states

For every screen or list that can be empty: | Screen/component | What is empty | Current empty state treatment | Has illustration/guidance? | Has call to action? | Rating |

Rating: OK = empty state is handled well; PARTIAL = present but incomplete; MISSING = no empty state handling.

Check:

  • Does the empty state explain why it is empty?
  • Does it tell the user what to do next?
  • Does it look intentional or broken?

2) Loading states

For every async operation: | Operation | Current loading indicator | Location | Blocks interaction? | Has timeout handling? |

Check:

  • Is there a loading indicator at all?
  • Does it appear immediately or after a delay?
  • Can the user still interact with other parts of the page?
  • What happens if loading takes more than 5 seconds? 30 seconds?
  • Is there a way to cancel?

3) Validation errors

For every form or input: | Form/input | Validation rules | When validation runs | Error message | Error placement | Recovery path |

Check:

  • Are errors shown inline or as a summary?
  • Do errors appear on submit or on blur?
  • Are error messages specific ("Email must include @") or vague ("Invalid input")?
  • Does the form preserve entered data on error?
  • Are required fields marked before the user submits?

4) Permission and authorization failures

For every protected action or screen: | Action/screen | Required permission | What happens without permission | User feedback |

Check:

  • Are unauthorized actions hidden, disabled, or shown with an error?
  • Does the error explain what permission is needed?
  • Does it explain how to get that permission?
  • Can users accidentally navigate to pages they cannot use?

5) Network and connectivity failures

| Operation | Behavior on network failure | Retry available? | Data preserved? | Offline fallback? |

Check:

  • What happens if the API returns 500?
  • What happens if the request times out?
  • Is there a global error boundary or does each component handle its own?
  • Are partial saves possible?

6) Partial completion

| Flow | Can be partially completed? | What happens if abandoned? | Can be resumed? | Is progress saved? |

Check:

  • Multi-step forms: what happens if the user closes mid-flow?
  • Drafts: are they auto-saved?
  • Bulk operations: what happens if some items succeed and some fail?

7) Undo, cancel, and recovery

| Action | Reversible? | Undo mechanism | Confirmation required? | Time limit on undo? |

Check:

  • Destructive actions: is there a confirmation dialog?
  • Delete: is it soft delete or permanent?
  • Cancel during a flow: does it save progress or discard everything?
  • Back button: does it work as expected at every step?

8) Concurrent and conflict states

| Scenario | Current handling | User feedback |

Check:

  • What happens if two users edit the same thing?
  • What happens if the data changes while the user is viewing it?
  • Are there stale state problems?

Summary table

CategoryItems auditedHandled wellPartially handledNot handledCritical gaps

Synthesis

End with:

  • top 5 most likely failure scenarios users will encounter
  • top 5 missing or broken failure states
  • patterns: which failure category is systematically ignored?
  • quickest wins: failure states that are easy to fix
  • structural issues: failure handling that requires architectural changes

Quality bar

  • Reference actual components, error boundaries, API error handling, and UI states.
  • Every happy path implies at least 3-5 failure paths. If you find fewer, look harder.
  • A good failure path audit should make the developer uncomfortable. That is the point.

Deliverable style

  • Clear headings
  • Markdown tables
  • Mermaid diagrams where useful
  • Direct language
  • No filler

Download now (ZIP)

Does 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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.