Skip to main content

Task Accountability Skill

Ensure every task has clear ownership, evidence of completion, and traceable outcomes for complete accountability in agentic workflows.

Purpose​

This skill establishes accountability chains for AI agent work. It ensures that every task can be traced to its origin, execution, and outcome - answering "who did what, when, and why" for any piece of work.

When to Use​

  • Before starting any task (establish ownership)
  • After completing any task (record evidence)
  • During quality gate verification
  • For audit and compliance reviews
  • When debugging or investigating issues

Accountability Dimensions​

1. Ownership Accountability​

Every task must have clear ownership:

**Task:** E.1.1 - Full signup → payment → activation E2E test
**Owner:** testing-specialist (Claude Code agent)
**Initiated By:** User request (2026-01-01 22:45:00Z)
**Context:** PILOT-PARALLEL-EXECUTION-PLAN.md, Track E.1
**Session:** 9ddd8c22-28cb-4b1e-9174-676bbece6de6

2. Evidence Accountability​

Every completion claim must have evidence:

**Completion Evidence for E.1.1:**

✅ File Created:
   └─ backend/tests/e2e/test_signup_activation_flow.py
   └─ 450 lines, 18,432 bytes
   └─ SHA-256: a7f3b2c...

✅ Tests Implemented:
   └─ 10 test methods across 3 classes
   └─ TestSignupToActivationFlow (4 tests)
   └─ TestPaymentWebhookReliability (3 tests)
   └─ TestMultiTenantActivationFlow (3 tests)

✅ Plan Updated:
   └─ PILOT-PARALLEL-EXECUTION-PLAN.md
   └─ Line 312: [x] E.1.1 - Full signup...
   └─ Diff: https://github.com/.../commit/abc123

✅ Session Logged:
   └─ SESSION-LOG-2026-01-01.md
   └─ Entry timestamp: 2026-01-01T23:00:00Z

3. Outcome Accountability​

Track the actual outcomes vs. expected:

**Expected Outcomes:**
- [ ] E2E test file created
- [ ] Coverage: signup, payment, activation
- [ ] At least 5 test methods
- [ ] Fixtures for api_client, tenant, user

**Actual Outcomes:**
- [x] E2E test file created ✅
- [x] Coverage: signup, payment, activation ✅
- [x] 10 test methods (exceeded target) ✅
- [x] Fixtures: api_client, tenant, user, mock_stripe, mock_redis ✅

**Outcome Score:** 100% (4/4 criteria met, 2 exceeded)

4. Chain of Custody​

Track the full lifecycle of work:

Task E.1.1 Chain of Custody
═══════════════════════════════════════════════════════════════
2026-01-01 22:45:00Z │ CREATED   │ User request via conversation
2026-01-01 22:45:30Z │ ASSIGNED  │ testing-specialist selected
2026-01-01 22:46:00Z │ STARTED   │ Agent began execution
2026-01-01 23:00:00Z │ COMPLETED │ File written, tests created
2026-01-01 23:00:30Z │ VERIFIED  │ Plan updated, log entry added
2026-01-01 23:01:00Z │ ACCEPTED  │ Quality gate passed

Quality Gates​

Gate 1: Pre-Execution Checklist​

Before starting any task:

  • Task ID assigned
  • Owner agent identified
  • Expected outcomes defined
  • Context sources documented
  • Session ID captured

Gate 2: Post-Execution Checklist​

After completing any task:

  • All expected outcomes met
  • Files created/modified documented
  • Plan updated with completion
  • Session log entry added
  • Evidence collected

Gate 3: Verification Checklist​

Before marking task complete:

  • Evidence matches claims
  • Quality criteria met
  • No regressions introduced
  • Documentation updated

Anti-Patterns to Avoid​

Anti-PatternProblemSolution
Orphan tasksNo traceable originAlways register task source
Claimed without evidenceUnverifiable completionRequire evidence collection
Single-point ownershipBus factor riskDocument delegation chain
Silent modificationsHidden changesLog all file operations

Accountability Metrics​

MetricTargetMeasurement
Task registration100%All tasks have registration record
Evidence coverage100%All completions have evidence
Outcome verification100%All outcomes verified against criteria
Chain completeness100%Full lifecycle documented

Success Output​

When this skill is successfully applied, output:

✅ SKILL COMPLETE: task-accountability

Completed:
- [x] Task E.1.1 registered with owner, context, session ID
- [x] Completion evidence collected (file paths, sizes, checksums)
- [x] Outcome verification completed (4/4 criteria met)
- [x] Chain of custody documented (6 lifecycle events)
- [x] Plan updated with completion status

Outputs:
- Task registration record with metadata
- Evidence package (files created, tests implemented, plan updates)
- Outcome scorecard (expected vs. actual)
- Chain of custody log (timestamps, state transitions)
- Session log entry

Completion Checklist​

Before marking this skill as complete, verify:

  • Task ID assigned and registered
  • Owner agent identified and documented
  • Expected outcomes defined with clear criteria
  • All expected outcomes met or exceeded
  • File operations documented (paths, sizes, checksums)
  • Plan updated with completion marker
  • Session log entry added
  • Evidence collected and verifiable
  • Quality gates passed (no regressions)

Failure Indicators​

This skill has FAILED if:

  • ❌ Task has no traceable origin or initiator
  • ❌ Claims of completion without evidence
  • ❌ Expected outcomes not met but marked complete
  • ❌ File modifications undocumented
  • ❌ Plan not updated with completion status
  • ❌ No session log entry
  • ❌ Chain of custody incomplete or missing
  • ❌ Quality criteria not verified

When NOT to Use​

Do NOT use this skill when:

  • Exploratory work with no deliverable - use research log instead
  • Trivial tasks (<5 min, no files changed) - overhead not justified
  • Prototyping or experiments - lighter tracking sufficient
  • Tasks delegated to human team members - use project management tools instead
  • Real-time debugging sessions - post-session summary acceptable
  • No compliance or audit requirements - simpler tracking may suffice

Use alternatives instead:

  • Exploration → Research log with findings
  • Trivial tasks → Session notes only
  • Prototyping → Lightweight task list
  • Human delegation → JIRA, Linear, etc.

Anti-Patterns (Avoid)​

Anti-PatternProblemSolution
Orphan tasksNo traceable originAlways register task source and initiator
Claimed without evidenceUnverifiable completionCollect file paths, checksums, diffs
Single-point ownershipBus factor riskDocument delegation chain and context
Silent modificationsHidden changes, audit failuresLog all file operations with timestamps
Skipping quality gatesRegressions undetectedVerify no breaking changes before marking complete
Incomplete chain of custodyAudit trail breaksDocument all lifecycle state transitions
No outcome verificationSuccess unclearDefine and verify expected outcomes

Principles​

This skill embodies:

  • #5 Eliminate Ambiguity - Clear ownership, evidence, and completion criteria
  • #6 Clear, Understandable, Explainable - Full transparency of who did what, when, why
  • #8 No Assumptions - Evidence-based verification, not trust-based claims
  • #9 Transparency - Complete audit trail from initiation to completion
  • #10 Traceability - Every action traceable to origin and outcome

Full Standard: CODITECT-STANDARD-AUTOMATION.md

  • Commands: /quality-gate, /audit-trail
  • Skills: process-transparency, audit-logging
  • Hooks: task-completion, quality-gate-check
  • Agents: quality-gate-enforcer

Skill Version: 1.0.0 Created: 2026-01-02 Author: CODITECT Process Refinement