self-improving-negotiation

Logs redacted negotiation learnings (concession leaks, BATNA gaps, framing misses, objections, agreement risk). Optional project-scoped reminder hooks. extract-skill.sh is dry-run by default and writes a SKILL.md scaffold only with --write after explicit user approval. Does not accept terms, set pricing, or approve deals. Use when negotiations stall, concessions exceed guardrails, terms are ambiguous, or recurring bargaining patterns emerge.

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

Situation	Action
Concession made without reciprocal value	Log NEG in .learnings/NEGOTIATION_ISSUES.md category concession_leak
BATNA undefined or weak	Log NEG category batna_weakness in preparation
Opening anchor failed or drifted	Log LRN category anchor_error
Same objection repeated across deals	Log LRN category objection_handling_gap
Escalation executed to wrong owner/timing	Log NEG category escalation_misalignment
Redline unresolved at close stage	Log NEG category agreement_risk
Value reframing improved movement	Log LRN category value_articulation_gap
Clause ambiguity discovered	Log NEG category agreement_risk in contract_terms
Workflow/tooling needed	Log FEAT in .learnings/FEATURE_REQUESTS.md
Pattern repeats 3+ times	Link 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

Category	Use When
concession_leak	Concession granted without reciprocal value
batna_weakness	BATNA undefined, unrealistic, or not actionable
framing_miss	Value narrative fails to move negotiation
objection_handling_gap	Repeated objection lacks effective response
escalation_misalignment	Escalation timing/owner path is wrong
value_articulation_gap	Team cannot connect terms to measurable value
anchor_error	Anchor too weak/strong or abandoned prematurely
agreement_risk	Final terms contain ambiguity or unresolved risk

Area Tags

Area	Scope
preparation	BATNA setup, approval limits, stakeholder map
discovery	Interests, constraints, hidden blockers
framing	Value framing and anchor setup
bargaining	Offer-counteroffer strategy
concessions	Give-get discipline and thresholds
closing	Final package shaping and confirmation
stakeholder_alignment	Internal/external decision alignment
escalation	Escalation path and timing
contract_terms	Clauses, redlines, obligations, ambiguity
follow_up	Recaps, 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

Priority	When to Use	Example
critical	High-impact unresolved risk or unauthorized concession	Liability term unresolved at close; major concession without approval
high	Recurring pattern with material impact	BATNA absent in enterprise cycles; repeated deadlocks
medium	Single-cycle issue with contained impact	One objection flow underperforming
low	Minor improvement or documentation cleanup	Non-critical wording updates

Promotion Targets

When patterns are validated and reusable, promote to:

Target	What to Promote
Negotiation playbooks	End-to-end tactics across stages
Objection libraries	Reusable objection-response flows
Concession guardrails	Give-get rules and approval thresholds
BATNA checklists	Preflight criteria and walk-away readiness
Deal review templates	Structured 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

Criterion	Requirement
Recurring	Appears in 3+ negotiations
Verified	Improvement proven in outcomes
Non-obvious	Practical experience required
Broad	Useful across contexts
Requested	User/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

Agent	Activation	Detection
Claude Code	Hook scripts	Automatic
Codex CLI	Hook scripts	Automatic
GitHub Copilot	Instruction guidance	Manual
OpenClaw	Workspace + hooks	Automatic 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.