Self-Improving Negotiation Skill#

Log negotiation learnings, negotiation issues, and feature requests to markdown files for continuous improvement. Capture preparation gaps, framing misses, concession leakage, weak BATNA posture, recurring objections, escalation mistakes, contract-term ambiguity, and agreement quality risk.

Optional extras (off by default):

• project-scoped reminder hooks
• extract-skill.sh (dry-run unless --write and the user asked in this session)

Promote validated, redacted patterns into playbooks, objection libraries, concession guardrails, BATNA checklists, or deal review templates after a reviewed diff and explicit approval.

Safety Posture#

Default behavior is local markdown logging and reminders.

It does not:

• auto-accept terms
• commit pricing
• execute legal/financial approvals
• finalize agreements
• run network calls unless you choose ClawHub install or git clone
• write skill files unless you run extract-skill.sh --write

Always require explicit human approval for high-impact concessions and for final terms.

Sensitive data#

Do not log BATNA values, reservation prices, discount floors, confidential customer terms, legal advice, approval tokens, or raw transcripts. Use redacted process-level summaries (what went wrong, what guardrail to add). Keep .learnings/ local or access-controlled; do not commit it until a human has reviewed entries.

First-Use Initialisation#

Before logging, ensure .learnings/ exists in project/workspace root. Create missing files only:

bash

mkdir -p .learnings
[ -f .learnings/LEARNINGS.md ] || printf "# Negotiation Learnings\n\nNegotiation strategy insights, framing improvements, BATNA lessons, concession tactics, objection handling patterns, escalation learnings, and agreement quality improvements.\n\n**Categories**: concession_leak | batna_weakness | framing_miss | objection_handling_gap | escalation_misalignment | value_articulation_gap | anchor_error | agreement_risk\n**Areas**: preparation | discovery | framing | bargaining | concessions | closing | stakeholder_alignment | escalation | contract_terms | follow_up\n\n---\n" > .learnings/LEARNINGS.md
[ -f .learnings/NEGOTIATION_ISSUES.md ] || printf "# Negotiation Issues Log\n\nNegotiation failures, concession breakdowns, unresolved redlines, escalation misalignment, and agreement quality risks.\n\n---\n" > .learnings/NEGOTIATION_ISSUES.md
[ -f .learnings/FEATURE_REQUESTS.md ] || printf "# Feature Requests\n\nNegotiation tooling, preparation automation, objection support, BATNA planning, concession controls, and agreement-risk analysis capabilities.\n\n---\n" > .learnings/FEATURE_REQUESTS.md

Never overwrite existing files.

Quick Reference#

SituationActionConcession made without reciprocal valueLog NEG in .learnings/NEGOTIATION_ISSUES.md category concession_leakBATNA undefined or weakLog NEG category batna_weakness in preparationOpening anchor failed or driftedLog LRN category anchor_errorSame objection repeated across dealsLog LRN category objection_handling_gapEscalation executed to wrong owner/timingLog NEG category escalation_misalignmentRedline unresolved at close stageLog NEG category agreement_riskValue reframing improved movementLog LRN category value_articulation_gapClause ambiguity discoveredLog NEG category agreement_risk in contract_termsWorkflow/tooling neededLog FEAT in .learnings/FEATURE_REQUESTS.mdPattern repeats 3+ timesLink entries; propose a redacted playbook patch after approval

OpenClaw Setup (Recommended)#

OpenClaw supports workspace-based skill loading and hook reminders.

Install#

Prefer the registry install (no extra clone). This uses the network to ClawHub:

bash

clawdhub install self-improving-negotiation

Manual clone is optional network access to GitHub. Use only if you trust that repository:

bash

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

Workspace Structure#

text

~/.openclaw/workspace/
├── AGENTS.md
├── SOUL.md
├── TOOLS.md
├── MEMORY.md
├── memory/
│   └── YYYY-MM-DD.md
└── .learnings/
    ├── LEARNINGS.md
    ├── NEGOTIATION_ISSUES.md
    └── FEATURE_REQUESTS.md

Create learning files in OpenClaw workspace#

bash

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

Then create:

• LEARNINGS.md
• NEGOTIATION_ISSUES.md
• FEATURE_REQUESTS.md

Optional Hook#

Do not install into ~/.openclaw/hooks/ (user-global). If reminders are needed, keep hooks in this workspace only:

bash

mkdir -p .openclaw/hooks
cp -r hooks/openclaw .openclaw/hooks/self-improving-negotiation

See references/openclaw-integration.md.

---

Generic Setup (Other Agents)#

For Claude Code, Codex CLI, Copilot, or other assistants:

bash

mkdir -p .learnings

Create the same 3 files in .learnings/.

Suggested instruction block#

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

When negotiation patterns are discovered:

1. Log a redacted summary to .learnings/LEARNINGS.md, .learnings/NEGOTIATION_ISSUES.md, or .learnings/FEATURE_REQUESTS.md
2. Review recurring items weekly
3. After a reviewed diff and explicit approval, promote reusable patterns to playbooks, objection libraries, concession guardrails, BATNA checklists, or deal review templates

Entry Types#

• LRN = learning
• NEG = negotiation issue
• FEAT = feature request

Logging Formats#

Learning Entry [LRN-YYYYMMDD-XXX]#

Append to .learnings/LEARNINGS.md:

markdown

## [LRN-YYYYMMDD-XXX] category

**Logged**: ISO-8601 timestamp
**Priority**: low | medium | high | critical
**Status**: pending
**Area**: preparation | discovery | framing | bargaining | concessions | closing | stakeholder_alignment | escalation | contract_terms | follow_up

### Summary
One-line negotiation learning.

### Details
What happened, why it mattered, and what to improve.

### Negotiation Context

**Trigger / Objection:**
> Quote or paraphrase

**Response Used:**
> Action, phrase, offer, or decision

**Outcome:**
advanced | stalled | rejected | escalated | closed

### Suggested Action
Concrete change to framing, concession strategy, BATNA prep, or escalation.

### Metadata
- Source: transcript_review | negotiation_debrief | legal_review | pricing_review | post_mortem
- Negotiation Type: procurement | renewal | enterprise_new_logo | partner | vendor | internal_budget
- Counterparty Role: procurement | legal | finance | exec | ops | technical
- Related Issues: NEG-YYYYMMDD-XXX
- Tags: tag1, tag2
- See Also: LRN-YYYYMMDD-XXX
- Pattern-Key: framing.anchor_drift (optional)
- Recurrence-Count: 1 (optional)
- First-Seen: YYYY-MM-DD (optional)
- Last-Seen: YYYY-MM-DD (optional)

---

Negotiation Issue Entry [NEG-YYYYMMDD-XXX]#

Append to .learnings/NEGOTIATION_ISSUES.md:

markdown

## [NEG-YYYYMMDD-XXX] issue_type

**Logged**: ISO-8601 timestamp
**Priority**: high
**Status**: pending
**Area**: preparation | discovery | framing | bargaining | concessions | closing | stakeholder_alignment | escalation | contract_terms | follow_up

### Summary
Brief description of failure/risk.

### Context
- **Stage**: current stage
- **Counterparty**: role/team
- **Cycle Day**: days from first commercial exchange

### What Happened
Sequence of events and where the issue surfaced.

### Root Cause
Preparation gap, weak BATNA, framing miss, concession error, approval gap, or term ambiguity.

### Impact
- Margin/pricing impact range
- Timeline impact
- Business/legal risk impact

### Prevention
Guardrails, scripts, checklists, escalation path changes.

### Metadata
- Trigger: deadlock | repeated_objection | scope_creep | concession_exceeded_threshold | batna_not_defined | approval_missing | term_ambiguity | redline_unresolved
- Related Files: path/to/debrief.md
- See Also: NEG-YYYYMMDD-XXX

---

Feature Request Entry [FEAT-YYYYMMDD-XXX]#

Append to .learnings/FEATURE_REQUESTS.md:

markdown

## [FEAT-YYYYMMDD-XXX] capability_name

**Logged**: ISO-8601 timestamp
**Priority**: medium
**Status**: pending
**Area**: preparation | discovery | framing | bargaining | concessions | closing | stakeholder_alignment | escalation | contract_terms | follow_up

### Requested Capability
Tooling or workflow enhancement needed.

### User Context
Why this is needed and what friction it removes.

### Complexity Estimate
simple | medium | complex

### Suggested Implementation
Practical implementation suggestion.

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

---

Categories#

CategoryUse Whenconcession_leakConcession granted without reciprocal valuebatna_weaknessBATNA undefined, unrealistic, or not actionableframing_missValue narrative fails to move negotiationobjection_handling_gapRepeated objection lacks effective responseescalation_misalignmentEscalation timing/owner path is wrongvalue_articulation_gapTeam cannot connect terms to measurable valueanchor_errorAnchor too weak/strong or abandoned prematurelyagreement_riskFinal terms contain ambiguity or unresolved risk

Area Tags#

AreaScopepreparationBATNA setup, approval limits, stakeholder mapdiscoveryInterests, constraints, hidden blockersframingValue framing and anchor setupbargainingOffer-counteroffer strategyconcessionsGive-get discipline and thresholdsclosingFinal package shaping and confirmationstakeholder_alignmentInternal/external decision alignmentescalationEscalation path and timingcontract_termsClauses, redlines, obligations, ambiguityfollow_upRecaps, unresolved items, next-round prep

ID Generation#

Format: TYPE-YYYYMMDD-XXX

• TYPE: LRN | NEG | FEAT
• YYYYMMDD: date
• XXX: sequential ( 001) or random ( A7B)

Examples:

• LRN-20260413-001
• NEG-20260413-A3F
• FEAT-20260413-002

Resolve Workflow#

When resolved:

1. Update **Status**: pending -> **Status**: resolved
2. Add resolution block:

markdown

### Resolution
- **Resolved**: 2026-04-13T14:30:00Z
- **Action Taken**: Added concession guardrail and approval checklist
- **Notes**: Applied successfully in two subsequent negotiations

Additional statuses:

• in_progress
• wont_fix
• promoted
• promoted_to_skill

Detection Triggers#

Log when these appear:

Deadlock#

• No movement after 2+ rounds
• Repeated counters without new rationale
• Decision timeline slips with no clear blocker

Repeated Objections#

• Same objection in 3+ negotiations
• Existing response script fails repeatedly
• Objection appears across multiple stakeholder roles

Scope Creep#

• New obligations added after pricing alignment
• Added deliverables without compensating terms
• Late-stage requirement expansion

Concession Exceeded Threshold#

• Discount/term concession above approved band
• Multiple micro-concessions exceed policy combined
• Concession offered before reciprocal commitment

BATNA Not Defined#

• No walk-away threshold
• No fallback path
• BATNA not costed/timed

Approval Missing#

• Concession offered without required approval
• Decision made on verbal approval only
• Approval owner unclear late in cycle

Term Ambiguity#

• Clause supports conflicting interpretations
• Undefined SLA/penalty terms
• Renewal/termination trigger unclear

Redline Unresolved#

• Open legal comments at close stage
• Internal and external redline versions diverge
• "To be clarified later" language remains

Priority Guidelines#

PriorityWhen to UseExamplecriticalHigh-impact unresolved risk or unauthorized concessionLiability term unresolved at close; major concession without approvalhighRecurring pattern with material impactBATNA absent in enterprise cycles; repeated deadlocksmediumSingle-cycle issue with contained impactOne objection flow underperforminglowMinor improvement or documentation cleanupNon-critical wording updates

Promotion Targets#

When patterns are validated and reusable, promote to:

TargetWhat to PromoteNegotiation playbooksEnd-to-end tactics across stagesObjection librariesReusable objection-response flowsConcession guardrailsGive-get rules and approval thresholdsBATNA checklistsPreflight criteria and walk-away readinessDeal review templatesStructured post-negotiation reviews

Promotion Criteria#

• Pattern recurs at least 3 times
• Countermeasure verified in real outcomes
• Useful across contexts (not one-off)
• Non-obvious and operationally important

Promotion Workflow#

1. Distill the learning into a concise, redacted asset.
2. Show a reviewed diff for the target document.
3. Apply only after explicit user approval, then update the source entry:
   • **Status**: promoted
   • **Promoted**: negotiation_playbook (or relevant target)

Recurring Pattern Detection#

Before creating a new entry:

1. Search existing patterns: bash

   rg "Pattern-Key|concession_leak|batna_weakness|anchor_error|agreement_risk" .learnings/
   
   

2. If related entry exists:

   • add See Also
   • increase Recurrence-Count
   • update Last-Seen

3. Raise priority if recurrence continues.

4. Identify systemic fix:

   • framing issue -> playbook update
   • objection issue -> objection library
   • concession issue -> guardrail policy
   • BATNA issue -> checklist hard gate
   • term issue -> contract review template

Simplify/Harden Feed#

When importing simplify-and-harden insights:

1. Use Pattern-Key as dedupe key.

2. Search for key: bash

   rg "Pattern-Key: <key>" .learnings/LEARNINGS.md
   
   

3. Existing key:

   • increment recurrence
   • update dates
   • link related entries

4. New key:

   • create LRN entry
   • set source to simplify-and-harden

Promotion threshold suggestion:

• recurrence >= 3
• appears in 2+ negotiation contexts
• appears within 90 days

Periodic Review#

When to review#

• Before high-stakes negotiation rounds
• After deadlock or failed escalation
• Weekly for active deals
• Monthly for pattern consolidation
• Before quarter-end closes

Quick checks#

bash

# Pending entries
rg "Status\\*\\*: pending" .learnings/*.md | wc -l

# High and critical issues
rg -n "Priority\\*\\*: high|Priority\\*\\*: critical" .learnings/NEGOTIATION_ISSUES.md

# BATNA issues
rg -n "batna_weakness|BATNA" .learnings/*.md

# Concession leaks
rg -n "concession_leak|concession" .learnings/*.md

# Agreement risk
rg -n "agreement_risk|term_ambiguity|redline_unresolved" .learnings/NEGOTIATION_ISSUES.md

Review actions#

• Resolve addressed issues
• Promote validated patterns
• Merge duplicates with links
• Tighten concession thresholds
• Refresh BATNA checklist
• Update escalation matrix

Best Practices#

1. Define BATNA, reservation point, and walk-away before first offer.
2. Anchor on value and risk reduction, not only price.
3. Trade every concession for reciprocal value.
4. Avoid stacking concessions in one round.
5. Keep approval ownership explicit before offering exceptions.
6. Treat ambiguous terms as active risk.
7. Capture objection wording precisely for reuse.
8. Escalate with clear decision owner and deadline.
9. Log outcomes the same day while context is fresh.
10. Promote recurring patterns after review.

Hook Integration (Opt-In)#

Use hooks to inject reminders and detect negotiation signals.

Hooks persist across sessions once installed. Keep them project-scoped. Do not install user-level or global hooks. Never use an empty matcher. PostToolUse inspects command output in-process; do not log raw output, secrets, or transcripts.

Example configuration#

json

{
  "hooks": {
    "UserPromptSubmit": [
      {
        "matcher": "counteroffer|concession|anchor|BATNA|walk-away|term.?sheet",
        "hooks": [
          { "type": "command", "command": "./skills/self-improving-negotiation/scripts/activator.sh" }
        ]
      }
    ],
    "PostToolUse": [
      {
        "matcher": "Bash",
        "hooks": [
          { "type": "command", "command": "./skills/self-improving-negotiation/scripts/error-detector.sh" }
        ]
      }
    ]
  }
}

Hook behavior is reminder-only; no execution of agreement or approval actions.

See references/hooks-setup.md.

Automatic Skill Extraction#

Extracted skills are untrusted until a human reviews the generated SKILL.md. Do not keep or publish an extracted skill without explicit user approval.

If a learning is stable and reusable and the user asked to extract a skill in this session, run the helper. Default is dry-run (no files written). Pass --write only after that explicit request.

Criteria#

CriterionRequirementRecurringAppears in 3+ negotiationsVerifiedImprovement proven in outcomesNon-obviousPractical experience requiredBroadUseful across contextsRequestedUser/team asked for standardization

Workflow#

1. Select candidate LRN or NEG.

2. Preview, then write only if the user approved: bash

   ./skills/self-improving-negotiation/scripts/extract-skill.sh skill-name
   ./skills/self-improving-negotiation/scripts/extract-skill.sh skill-name --write
   
   

3. Fill TODO placeholders.

4. Update source entry:

   • **Status**: promoted_to_skill
   • **Skill-Path**: skills/<skill-name>

5. Validate in fresh session.

Multi-Agent Support#

AgentActivationDetectionClaude CodeHook scriptsAutomaticCodex CLIHook scriptsAutomaticGitHub CopilotInstruction guidanceManualOpenClawWorkspace + hooksAutomatic reminders

Coordination rules:

• Use one canonical .learnings/ folder
• Use stable IDs ( LRN, NEG, FEAT)
• Add See Also cross-links and reviewer ownership before promotion

Agreement Quality Gate#

Before final recommendation, confirm:

• BATNA is still valid
• concessions are reciprocal
• approvals are documented
• redlines are resolved
• ambiguous clauses are removed

If any check fails, log NEG and pause final recommendation.

Gitignore Options#

gitignore

.learnings/

gitignore

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

Shared team mode: track .learnings/ only after a human has reviewed entries for pricing strategy, BATNA, legal, and customer-confidential content.

References#

• references/examples.md
• references/hooks-setup.md
• references/openclaw-integration.md
• Clone URL: https://github.com/jose-compu/self-improving-negotiation.git Skill, hooks, and scripts are reminder-only controls; they never approve concessions, accept terms, commit pricing, execute approvals, or finalize agreements. Human approval remains required for high-impact concessions and final terms.

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/negotiation/
• 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:

markdown

**Skill**: negotiation

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 (negotiation)#

Only trigger this skill automatically for negotiation signals such as:

• counteroffer|concession|anchor|walk-away|term sheet
• deal risk|stakeholder alignment|fallback option|BATNA
• explicit negotiation 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/negotiation/ in stackable mode.
• It may read other skill folders for cross-linking, but should not rewrite their entries.
• Standalone mode writes to this project's .learnings/*.md log files only.
• Stackable mode writes only to the namespaced folder above and must not rewrite other skills' log entries.
• Promotion into AGENTS.md, SOUL.md, TOOLS.md, MEMORY.md, rules, hooks, or generated skills is not a logging write. Show a reviewed diff and apply only after explicit user approval.