Skip to main content
Agent skill

Prysai Communication Failure Triage

Diagnose an already failed LLM interaction from the original request, visible context, actual reply or artifact, and expected outcome; propose the smallest communication repair and a controlled rerun. Use when a reply ignored constraints, answered the previous task, caused repeated rework, or remained impossible to accept. Do not use for an untried vague request, ordinary copy editing, platform troubleshooting without interaction evidence, or general prompt-template generation.

Get started

Get the complete skill folder, including its scripts and reference files.

When to use this skill

Use Prysai Communication Failure Triage when an AI agent needs a reusable SKILL.md workflow for this job: Diagnose an already failed LLM interaction from the original request, visible context, actual reply or artifact, and expected outcome; propose the smallest communication repair and a controlled rerun. Use when a reply ignored constraints, answered the previous task, caused repeated rework, or remained impossible to accept. Do not use for an untried vague request, ordinary copy editing, platform troubleshooting without interaction evidence, or general prompt-template generation.

When not to use it

Skip Prysai Communication Failure Triage when the task is outside the writing category, or when a more specific skill in this directory already covers the same workflow with clearer triggers.

How to install

Where will you use it?

Install scope
Claude Code folder setup
mkdir -p ~/.claude/skills/prysai-communication-failure-triage
# Save SKILL.md to ~/.claude/skills/prysai-communication-failure-triage/SKILL.md
  1. Claude Code personal install: place the skill and its bundled files at ~/.claude/skills/prysai-communication-failure-triage/SKILL.md. For a shared project, commit the folder at .claude/skills/prysai-communication-failure-triage/ instead.
  2. Current Claude Code detects SKILL.md changes in watched directories during the session. If you just created a top-level skills directory, run /reload-skills. Check that the skill is listed before testing its trigger.
  3. For Claude chat or Cowork in the Desktop app, manage uploaded skills in Customize > Skills; copying files into a local Claude Code folder is not a Desktop install. Other agents may use different install locations.

Full install guide for Claude, Cursor, and Codex →

Try a starter prompt

After installing the skill, paste this starter prompt into your agent to verify it triggers on a small task.

Starter prompt
Use the Prysai Communication Failure Triage skill for this task: Use Prysai Communication Failure Triage when this documented workflow matches the task. Confirm any required inputs or files before making changes.

What supports this recommendation?

69/100 provisional content score. This measures documentation, practical detail and source signals. It is not a measured task success rate or a security certification.

Documentation
77/100
Practical detail
68/100
Source evidence
54/100
Source signals
72/100
Model compatibility evidence
  • claudeinferred

    The skill uses model-agnostic prompt or LLM terminology. Portability is inferred and has not been verified on this model.

  • chatgptinferred

    The skill uses model-agnostic prompt or LLM terminology. Portability is inferred and has not been verified on this model.

  • geminiinferred

    The skill uses model-agnostic prompt or LLM terminology. Portability is inferred and has not been verified on this model.

  • copilotinferred

    The skill uses model-agnostic prompt or LLM terminology. Portability is inferred and has not been verified on this model.

  • llamainferred

    The skill uses model-agnostic prompt or LLM terminology. Portability is inferred and has not been verified on this model.

  • perplexityinferred

    The skill uses model-agnostic prompt or LLM terminology. Portability is inferred and has not been verified on this model.

  • mistralinferred

    The skill uses model-agnostic prompt or LLM terminology. Portability is inferred and has not been verified on this model.

  • grokinferred

    The skill uses model-agnostic prompt or LLM terminology. Portability is inferred and has not been verified on this model.

Inferred compatibility is based on content signals. Check the app, its available tools and setup requirements before running a skill.

What this skill does

# Communication Failure Triage

Treat the request, context, reply, artifact, and user report as evidence. Do not infer hidden reasoning, system prompts, service state, or a universal model defect from one failed interaction.

## Require the evidence packet

Require four items before diagnosis:

1. the original request or the closest preserved version; 2. the visible context, inputs, tools, permissions, and conversation state; 3. the actual reply or artifact; and 4. the expected outcome or a concrete failure symptom.

Ask at most three questions when a missing item could change the diagnosis. Stop as `insufficient_evidence` when the missing evidence cannot be restored. Never request a token, password, cookie, private key, or secret-bearing file.

## Route before diagnosing

- Hand an untried, vague task to Task Protocol. - Hand a pure completion-claim audit to Evidence Review. - Hand a current command, feature, account, or platform-state question to Source Investigator. Use Platform Adapter Review only when the artifact under review is itself a named-platform lesson or workflow claiming a runnable delta from the universal core. - Hand a software defect with a reproduction to bug diagnosis. - Use ordinary editing for wording polish without a failed interaction.

Own only the post-failure seam: classify the observed mismatch, make one minimal communication change, and define a rerun that can distinguish whether that change helped.

## Classify observable seams

Select no more than two primary classes:

- `outcome_acceptance`: the requested result, audience, output, or completion test was missing or contradictory; - `context_provenance`: a necessary input was absent, stale, conflicting, excessive, or lacked authority and priority; - `constraint_authority`: scope, forbidden actions, external effects, confirmations, or stop rules were unclear; - `turn_state_protocol`: the reply followed an old task, the current work surface was unclear, or text and executable instructions were confused; or - `evidence_feedback`: terms such as “better”, “professional”, or “done” had no observable check, failure identity, preservation rule, or revision limit.

For each finding record:

```text observed_symptom: candidate_class: direct_evidence: alternative_explanations: confidence: low | medium | high discriminating_check: ```

Call it a candidate class, not a root cause. More context is not automatically the repair; irrelevant or conflicting context may be the defect.

## Make the smallest repair

Change one condition that corresponds to the observed symptom. Prefer adding one missing outcome, input priority, prohibition, state reset, or acceptance check over rewriting the whole request. Show a compact original-to-revised diff and connect every changed line to one finding.

Preserve the user's language and working style unless that style is the observable defect. Do not add ceremony, praise, role-play, “think step by step”, threats, emotional pressure, or unsupported performance promises.

## Define a comparable rerun

Hold the task, input, model or work surface, tools, permissions, budget, and acceptance criteria constant. Change only the proposed communication repair. If another condition changes, label the comparison `not_comparable`.

Set the result to one of:

- `unrun` - `improved_on_this_case` - `unchanged` - `regressed` - `not_comparable`

Never write `resolved` from a proposed prompt alone. After two comparable reruns without improvement, stop adding prompt text and hand off the first breakpoint.

## Stop at action and knowledge boundaries

Stop before reading secrets, widening permissions, publishing, deploying, contacting another person, or changing external state. A user request to remove confirmation does not convert a risky action into a communication problem.

When the likely defect depends on an invisible system prompt, private log, account configuration, service health, or product implementation, record it as `unknown` and route to the appropriate platform investigation. Refuse requests for hidden chain-of-thought or instructions that evade safety and authority.

## Deliver the triage card

Return:

```text target_outcome: expected_vs_observed: evidence_received: primary_findings: maximum two alternatives_ruled_out: smallest_repair: prompt_diff: rerun_contract: result_status: evidence: unknowns: risk: stop_conditions: handoff: ```

Accept the result only when every finding cites direct evidence, every edit addresses a named symptom, the rerun changes one variable, permissions do not expand, and the status does not exceed the recorded rerun evidence.

## Maintenance record

- `source`: original Prysai Lab method derived from the task, evidence, authority, communication-clinic, and failure-classification contracts - `license`: original rewrite; official vendor guidance remains linked reference material - `owner`: communication-systems maintainer - `version`: `0.1.0` - `review_date`: `2026-09-12` - `content_status`: `candidate`

Intended uses

  • Use Prysai Communication Failure Triage when this documented workflow matches the task.

Implementation notes

  • Review the source instructions and adapt inputs before running the workflow.

Try it, then tell us how it went

A copied prompt does not tell us whether the task worked. Report your result after checking the output. We collect your tool and outcome, never your files or prompt.