self-improving-support（自我提升支持）

Self-Improving Support Skill

Log support-specific learnings, ticket issues, and feature requests to markdown files for continuous improvement. Captures resolution delays, misdiagnoses, escalation gaps, SLA breaches, knowledge base gaps, and customer churn signals. Important learnings get promoted to KB articles, troubleshooting trees, escalation matrices, or canned responses.

First-Use Initialisation

Before logging anything, ensure the .learnings/ directory and files exist in the project or workspace root. If any are missing, create them:

mkdir -p .learnings
[ -f .learnings/LEARNINGS.md ] || printf "# Support Learnings\
\
Resolution delays, misdiagnoses, escalation gaps, SLA breaches, knowledge gaps, and churn signals captured during support operations.\
\
**Categories**: resolution_delay | misdiagnosis | escalation_gap | knowledge_gap | sla_breach | customer_churn_signal\
**Areas**: triage | diagnosis | resolution | follow_up | documentation | escalation\
\
---\
" > .learnings/LEARNINGS.md
[ -f .learnings/TICKET_ISSUES.md ] || printf "# Ticket Issues Log\
\
Recurring ticket problems, resolution failures, SLA breaches, and escalation breakdowns.\
\
---\
" > .learnings/TICKET_ISSUES.md
[ -f .learnings/FEATURE_REQUESTS.md ] || printf "# Feature Requests\
\
Support tooling, automation, and workflow improvement requests.\
\
---\
" > .learnings/FEATURE_REQUESTS.md

Never overwrite existing files. This is a no-op if .learnings/ is already initialised.

Do not log customer PII, account credentials, or internal auth tokens. Use ticket IDs and anonymised references. Prefer short summaries over verbatim customer messages.

If you want automatic reminders, use the opt-in hook workflow described in Hook Integration.

Quick Reference

OpenClaw Setup (Recommended)

OpenClaw is the primary platform for this skill. It uses workspace-based prompt injection with automatic skill loading.

Installation

Via ClawdHub (recommended):

clawdhub install self-improving-support

Manual:

git clone https://github.com/jose-compu/self-improving-support.git ~/.openclaw/skills/self-improving-support

Workspace Structure

OpenClaw injects these files into every session:

~/.openclaw/workspace/
├── AGENTS.md          # Multi-agent workflows, delegation patterns
├── SOUL.md            # Behavioral guidelines, personality, principles
├── TOOLS.md           # Tool capabilities, integration gotchas
├── MEMORY.md          # Long-term memory (main session only)
├── memory/            # Daily memory files
│   └── YYYY-MM-DD.md
└── .learnings/        # This skill's log files
    ├── LEARNINGS.md
    ├── TICKET_ISSUES.md
    └── FEATURE_REQUESTS.md

Create Learning Files

mkdir -p ~/.openclaw/workspace/.learnings

Then create the log files (or copy from assets/):

• LEARNINGS.md — resolution delays, escalation gaps, knowledge gaps, churn signals
• TICKET_ISSUES.md — misdiagnoses, SLA breaches, escalation failures, repeat tickets
• FEATURE_REQUESTS.md — support tooling, automation, workflow improvements

Promotion Targets

When support learnings prove broadly applicable, promote them:

Optional: Enable Hook

For automatic reminders at session start:

cp -r hooks/openclaw ~/.openclaw/hooks/self-improving-support
openclaw hooks enable self-improving-support

See references/openclaw-integration.md for complete details.

Generic Setup (Other Agents)

For Claude Code, Codex, Copilot, or other agents, create .learnings/ in the project or workspace root:

mkdir -p .learnings

Create the files inline using the headers shown above.

Add reference to agent files

Add to AGENTS.md, CLAUDE.md, or .github/copilot-instructions.md:

Self-Improving Support Workflow

When support issues or patterns are discovered:

1. Log to .learnings/TICKET_ISSUES.md, LEARNINGS.md, or FEATURE_REQUESTS.md
2. Review and promote broadly applicable learnings to:

- KB articles — customer-facing resolution guides - Troubleshooting trees — step-by-step diagnosis flows - Escalation matrices — routing rules for engineering teams - Canned responses — pre-approved reply templates

Logging Format

Learning Entry [LRN-YYYYMMDD-XXX]

Append to .learnings/LEARNINGS.md:

## [LRN-YYYYMMDD-XXX] category

**Logged**: ISO-8601 timestamp
**Priority**: low | medium | high | critical
**Status**: pending
**Area**: triage | diagnosis | resolution | follow_up | documentation | escalation

### Summary
One-line description of the support insight

### Details
Full context: what support situation occurred, why the outcome was suboptimal,
what the correct triage, diagnosis, or resolution approach is.
Include ticket timeline and relevant customer interaction summary (anonymised).

### Resolution Steps

**What happened:**
Brief timeline of the support interaction and outcome.

**What should have happened:**
The correct workflow or diagnosis path.

### Suggested Action
Specific KB article, escalation rule, or process change to adopt

### Metadata
- Source: repeat_ticket | csat_drop | sla_timer | ticket_reopen | escalation_failure | kb_search_miss
- Channel: email | chat | phone | portal | api
- Product Area: billing | auth | api | dashboard | integrations | infrastructure
- Customer Tier: free | starter | professional | enterprise
- Related Tickets: TKT-12345, TKT-12346
- Tags: tag1, tag2
- See Also: LRN-20250110-001 (if related to existing entry)
- Pattern-Key: misdiagnosis.dns_vs_ssl | escalation.wrong_team (optional)
- Recurrence-Count: 1 (optional)
- First-Seen: 2025-01-15 (optional)
- Last-Seen: 2025-01-15 (optional)

---

Categories for learnings:

Ticket Issue Entry [TKT-YYYYMMDD-XXX]

Append to .learnings/TICKET_ISSUES.md:

## [TKT-YYYYMMDD-XXX] issue_type_or_name

**Logged**: ISO-8601 timestamp
**Priority**: high
**Status**: pending
**Area**: triage | diagnosis | resolution | follow_up | documentation | escalation

### Summary
Brief description of the ticket issue

### Ticket Timeline
| Time | Event |
|------|-------|
| T+0h | Ticket opened by customer |
| T+Xh | First response sent |
| T+Xh | Diagnosis attempted |
| T+Xh | Escalation / resolution |
| T+Xh | Ticket closed or reopened |

### Root Cause
What in the support process caused the issue. Was it a misdiagnosis,
wrong escalation path, missing KB article, or SLA timer oversight?

### Correct Approach
What the agent should have done — the right triage, diagnosis, and
resolution path for this type of ticket.

### Customer Impact
- SLA Status: met | breached (first_response | resolution | update)
- CSAT Impact: none | minor | significant
- Reopened: yes | no
- Escalated: yes (to whom) | no

### Prevention
How to prevent this ticket issue from recurring (KB update, escalation
rule change, training material, canned response).

### Context
- Trigger: repeat_ticket | csat_drop | sla_breach | ticket_reopen | escalation_failure
- Channel: email | chat | phone | portal
- Product Area: billing | auth | api | dashboard | integrations | infrastructure
- Customer Tier: free | starter | professional | enterprise

### Metadata
- Related Tickets: TKT-12345
- Related Files: path/to/runbook.md
- See Also: TKT-20250110-001 (if recurring pattern)

---

Feature Request Entry [FEAT-YYYYMMDD-XXX]

Append to .learnings/FEATURE_REQUESTS.md:

## [FEAT-YYYYMMDD-XXX] capability_name

**Logged**: ISO-8601 timestamp
**Priority**: medium
**Status**: pending
**Area**: triage | diagnosis | resolution | follow_up | documentation | escalation

### Requested Capability
What support tool, automation, or workflow improvement is needed

### User Context
Why it's needed, what support workflow it improves, what ticket pattern it addresses

### Complexity Estimate
simple | medium | complex

### Suggested Implementation
How this could be built: Zendesk macro, Intercom workflow, Slack bot, KB template,
escalation rule, auto-categorizer, SLA monitor, CSAT survey trigger

### Metadata
- Frequency: first_time | recurring
- Related Features: existing_tool_or_workflow

---

ID Generation

Format: TYPE-YYYYMMDD-XXX

• TYPE: LRN (learning), TKT (ticket issue), FEAT (feature request)
• YYYYMMDD: Current date
• XXX: Sequential number or random 3 chars (e.g., 001, A7B)

Examples: LRN-20250415-001, TKT-20250415-A3F, FEAT-20250415-002

Resolving Entries

When an issue is addressed, update the entry:

1. Change **Status**: pending → **Status**: resolved
2. Add resolution block after Metadata:

### Resolution
- **Resolved**: 2025-01-16T09:00:00Z
- **Action Taken**: Updated KB article / revised escalation matrix / created canned response
- **Notes**: Added troubleshooting tree for DNS resolution failures

Other status values:

• in_progress — Actively being investigated or remediated
• wont_fix — Decided not to address (add reason in Resolution notes)
• promoted — Elevated to KB article, troubleshooting tree, or escalation matrix
• promoted_to_skill — Extracted as a reusable skill

Detection Triggers

Automatically log when you encounter:

Repeat Ticket Patterns (→ ticket issue with repeat_ticket trigger):

• Same customer opens ticket for same issue within 30 days
• Multiple customers report identical symptoms in same week
• Ticket references a previous ticket number
• Customer says "I already contacted you about this"

Negative Sentiment in Customer Messages (→ learning with customer_churn_signal):

• Explicit frustration: "This is unacceptable", "I've been waiting for days"
• Cancellation language: "cancel my subscription", "looking for alternatives"
• Escalation demand: "Let me speak to a manager", "This needs to be escalated"
• Disappointment: "I expected better", "This used to work fine"

SLA Timer Approaching (→ ticket issue with sla_breach trigger):

• First response SLA at 80% elapsed time
• Resolution SLA at 70% elapsed time
• Update cadence SLA missed (no update in promised interval)
• P1/P2 ticket without engineer assignment within threshold

Ticket Reopened (→ ticket issue with ticket_reopen trigger):

• Customer reopens within 7 days of closure
• Auto-reopen from monitoring alert on same issue
• Customer replies "This didn't fix my problem"
• Same error recurs in customer's system logs

CSAT Below Threshold (→ learning with customer_churn_signal):

• Individual ticket CSAT ≤ 2/5
• Rolling 7-day CSAT average drops below target
• Customer leaves negative free-text feedback
• NPS detractor response from support interaction

Knowledge Base Search Failures (→ learning with knowledge_gap):

• Agent searches KB and finds no relevant article
• Customer self-service search returns zero results
• Agent creates ticket for issue that should have a KB article
• Multiple agents ask the same question in internal channels

Priority Guidelines

Area Tags

Use to filter learnings by support domain:

Promoting to Permanent Support Standards

When a learning is broadly applicable (not a one-off ticket), promote it to permanent support resources.

When to Promote

• Same ticket issue recurs across multiple customers or channels
• Misdiagnosis pattern found in 3+ resolved tickets
• Knowledge gap affects multiple agents independently
• Resolution workflow would save significant handle time if documented

Promotion Targets

How to Promote

1. Distill the learning into a concise KB article, triage rule, or response template
2. Add to appropriate target (KB, escalation matrix, canned response library)
3. Update original entry:

- Change **Status**: pending → **Status**: promoted - Add **Promoted**: KB article (or troubleshooting tree, escalation matrix, canned response)

Promotion Examples

Learning → KB article:

Three agents misdiagnosed "502 Bad Gateway" as server-side when the root cause was customer's WAF blocking our API callbacks.

Promoted to: KB article "502 Bad Gateway on API Callbacks" with symptom, common misdiagnosis, actual cause (customer WAF/firewall), and IP whitelist resolution steps.

Learning → Escalation rule:

Enterprise customer escalated to VP after three P2 tickets for the same SSO issue. Each time a different agent started from scratch with no linked history.

Promoted to: Escalation matrix rule — on 3rd ticket for same issue category from enterprise customer, auto-assign previous agent, link history, notify account manager.

Recurring Pattern Detection

If logging something similar to an existing entry:

1. Search first: grep -r "keyword" .learnings/
2. Link entries: Add **See Also**: TKT-20250110-001 in Metadata
3. Bump priority if issue keeps recurring
4. Consider systemic fix: Recurring support issues often indicate:

- Missing KB article (→ write and publish) - Unclear escalation path (→ update escalation matrix) - Product defect (→ file engineering ticket with linked support tickets) - Training gap (→ create internal training material)

Periodic Review

Review .learnings/ at natural breakpoints:

When to Review

• Before starting a shift or taking over a ticket queue
• After a major incident or outage
• When the same ticket category spikes
• Weekly during support retrospectives

Quick Status Check

grep -h "Status\*\*: pending" .learnings/*.md | wc -l
grep -B5 "Priority\*\*: high" .learnings/TICKET_ISSUES.md | grep "^## \["
grep -B2 "sla_breach" .learnings/TICKET_ISSUES.md | grep "^## \["
grep -B2 "knowledge_gap" .learnings/LEARNINGS.md | grep "^## \["

Review Actions

• Resolve addressed ticket issues
• Promote recurring patterns to KB articles
• Link related entries across files
• Extract reusable resolutions as canned responses or troubleshooting trees

Simplify & Harden Feed

Ingest recurring support patterns from simplify-and-harden into KB articles or escalation rules.

1. For each candidate, use pattern_key as the dedupe key.
2. Search .learnings/LEARNINGS.md for existing entry: grep -n "Pattern-Key: <key>" .learnings/LEARNINGS.md
3. If found: increment Recurrence-Count, update Last-Seen, add See Also links.
4. If not found: create new LRN-... entry with Source: simplify-and-harden.

Promotion threshold: Recurrence-Count >= 3, seen across 2+ customers or channels, within 30-day window. Targets: KB articles, troubleshooting trees, escalation matrices, canned responses, SOUL.md / AGENTS.md.

Hook Integration

Enable automatic reminders through agent hooks. This is opt-in.

Quick Setup (Claude Code / Codex)

Create .claude/settings.json in your project:

{
  "hooks": {
    "UserPromptSubmit": [{
      "matcher": "",
      "hooks": [{
        "type": "command",
        "command": "./skills/self-improving-support/scripts/activator.sh"
      }]
    }]
  }
}

This injects a support-focused learning evaluation reminder after each prompt (~50-100 tokens overhead).

Advanced Setup (With Issue Detection)

{
  "hooks": {
    "UserPromptSubmit": [{
      "matcher": "",
      "hooks": [{
        "type": "command",
        "command": "./skills/self-improving-support/scripts/activator.sh"
      }]
    }],
    "PostToolUse": [{
      "matcher": "Bash",
      "hooks": [{
        "type": "command",
        "command": "./skills/self-improving-support/scripts/error-detector.sh"
      }]
    }]
  }
}

Enable PostToolUse only if you want the hook to inspect command output for SLA breaches, escalation failures, and customer dissatisfaction signals.

Available Hook Scripts

See references/hooks-setup.md for detailed configuration and troubleshooting.

Automatic Skill Extraction

When a support learning is valuable enough to become a reusable skill, extract it.

Skill Extraction Criteria

Extraction Workflow

1. Identify candidate: Learning meets extraction criteria
2. Run helper (or create manually):

   ./skills/self-improving-support/scripts/extract-skill.sh skill-name --dry-run
   ./skills/self-improving-support/scripts/extract-skill.sh skill-name

1. Customize SKILL.md: Fill in template with support-specific content
2. Update learning: Set status to promoted_to_skill, add Skill-Path
3. Verify: Read skill in fresh session to ensure it's self-contained

Extraction Detection Triggers

In conversation: "This keeps happening to customers", "Save this resolution as a skill", "Every agent struggles with this", "We need a runbook for this".

In entries: Multiple See Also links, high priority + resolved, knowledge_gap or misdiagnosis with broad applicability, same Pattern-Key across customer tiers.

Multi-Agent Support

Best Practices

1. Acknowledge first, diagnose second — validate the customer's experience before troubleshooting
2. Diagnose before acting — confirm root cause before applying a fix; misdiagnosis wastes everyone's time
3. Always follow up — check back after resolution to confirm the fix holds and customer is satisfied
4. Document every resolution — if you had to research it, future agents will too; write the KB article
5. Log immediately — context fades fast; capture the ticket issue while details are fresh
6. Anonymise customer data — use ticket IDs and product areas, never names or account details
7. Link related tickets — patterns only emerge when individual tickets are connected
8. Escalate with context — include what you already tried, what you ruled out, and what you suspect
9. Promote aggressively — if the same issue appears three times, it needs a KB article or automation
10. Review before shift — check .learnings/ for recent patterns in your ticket queue

Gitignore Options

Keep learnings local (per-agent):

.learnings/

Track learnings in repo (team-wide): Don't add to .gitignore — learnings become shared knowledge.

Hybrid (track templates, ignore entries):

.learnings/*.md
!.learnings/.gitkeep

Stackability Contract (Standalone + Multi-Skill)

This skill is standalone-compatible and stackable with other self-improving skills.

Namespaced Logging (recommended for 2+ skills)

• Namespace for this skill: .learnings/support/
• Keep current standalone behavior if you prefer flat files.
• Optional shared index for all skills: .learnings/INDEX.md

Required Metadata

Every new entry must include:

**Skill**: support

Hook Arbitration (when 2+ skills are enabled)

• Use one dispatcher hook as the single entrypoint.
• Dispatcher responsibilities: route by matcher, dedupe repeated events, and rate-limit reminders.
• Suggested defaults: dedupe key = event + matcher + file + 5m_window; max 1 reminder per skill every 5 minutes.

Narrow Matcher Scope (support)

Only trigger this skill automatically for support signals such as:

• ticket|sla breach|escalation|incident update|resolution note
• repeat issue|handoff|kb gap|customer impact
• explicit support intent in user prompt

Cross-Skill Precedence

When guidance conflicts, apply:

1. security
2. engineering
3. coding
4. ai
5. user-explicit domain skill
6. meta as tie-breaker

Ownership Rules

• This skill writes only to .learnings/support/ in stackable mode.
• It may read other skill folders for cross-linking, but should not rewrite their entries.