Process Transparency Skill

Enable full visibility into AI agent decision-making, task execution, and progress tracking for complete process transparency.

Purpose

This skill ensures that all AI agent activities are visible, understandable, and traceable. It addresses the "black box" problem in agentic workflows by making every decision and action explicit.

When to Use

Starting any multi-step task execution
When users need to understand AI decision rationale
During complex orchestration workflows
For compliance and audit requirements
When debugging or improving agent performance

Transparency Dimensions

1. Decision Transparency

Make the "why" behind every decision visible:

**Decision Point:** Agent Selection
**Options Considered:**
1. testing-specialist (confidence: 0.92) - Best for E2E tests
2. codi-qa-specialist (confidence: 0.78) - Alternative for QA
3. general-purpose (confidence: 0.65) - Fallback option

**Selected:** testing-specialist
**Rationale:** Highest confidence match for "create E2E test suite" task
**Evidence:** Pattern match on keywords: "E2E", "test", "integration"

2. Progress Transparency

Real-time visibility into task progress:

Track E.1 - Integration Testing
═══════════════════════════════════════════════════════════════
E.1.1 ████████████████████ 100% ✅ Complete (2026-01-01 23:00)
E.1.2 ████████████████████ 100% ✅ Complete (2026-01-01 23:15)
E.1.3 ████████████████████ 100% ✅ Complete (2026-01-01 23:30)
E.1.4 ████████████████████ 100% ✅ Complete (2026-01-01 23:45)
E.1.5 ████████████████████ 100% ✅ Complete (2026-01-02 00:00)
───────────────────────────────────────────────────────────────
Overall: 100% | Time: 1h 15m | Tests: 70 | Lines: 2,150

3. Action Transparency

Log every action taken:

{
  "timestamp": "2026-01-02T00:30:00Z",
  "action": "file_write",
  "target": "backend/tests/e2e/test_signup_activation_flow.py",
  "bytes_written": 18432,
  "line_count": 450,
  "agent": "testing-specialist",
  "task_id": "E.1.1",
  "rationale": "Implementing full signup → payment → activation E2E test"
}

4. Context Transparency

Show what context influenced decisions:

**Context Sources Used:**
1. PILOT-PARALLEL-EXECUTION-PLAN.md (lines 45-120)
2. SESSION-LOG-2026-01-01.md (recent 50 entries)
3. /cxq "E2E testing patterns" (3 relevant results)
4. Previous session checkpoint (2025-12-31)

**Context Weight:** 85% from plan, 15% from patterns

Implementation Patterns

Pattern 1: Progressive Disclosure

# Summary View (Default)
✅ E.1.1 Complete | 10 tests | 450 lines

# Expanded View (On Request)
✅ E.1.1 - Full signup → payment → activation E2E test
   └─ Agent: testing-specialist
   └─ Duration: 15 minutes
   └─ Files: test_signup_activation_flow.py
   └─ Tests: 10 methods in 3 classes
   └─ Coverage: Registration, JWT, Cart, Checkout, Webhooks, Entitlements

# Full Detail View (Audit Mode)
[Complete action log with timestamps, decisions, and evidence]

Pattern 2: Structured Logging

Every significant action follows this format:

def log_transparent_action(action: str, details: dict) -> None:
    """Log action with full transparency metadata."""
    entry = {
        "timestamp": datetime.now(timezone.utc).isoformat(),
        "action": action,
        "agent": current_agent(),
        "task_id": current_task_id(),
        "details": details,
        "context_hash": hash_current_context(),
        "decision_rationale": get_decision_rationale()
    }
    append_to_session_log(entry)

Pattern 3: Decision Trees

For complex decisions, document the full tree:

Agent Selection for "Create E2E tests"
├── Keyword Analysis
│   ├── "E2E" → testing domain (weight: 0.4)
│   ├── "tests" → testing domain (weight: 0.3)
│   └── "activation" → backend domain (weight: 0.2)
├── Capability Match
│   ├── testing-specialist: 92% match
│   ├── codi-qa-specialist: 78% match
│   └── backend-developer: 45% match
├── Historical Performance
│   └── testing-specialist: 98% success on similar tasks
└── SELECTED: testing-specialist (confidence: 0.92)

Anti-Patterns to Avoid

Anti-Pattern	Problem	Solution
Silent failures	User unaware of issues	Always log and surface errors
Implicit decisions	"Black box" behavior	Document every decision point
Batch updates only	Stale progress info	Real-time streaming updates
Technical-only logs	Users can't understand	Include human-readable summaries

Integration Points

Component	Integration
`/pilot --dashboard`	Real-time progress display
`/audit-trail`	Complete action history
Session logs	Persistent record
Context snapshots	State transparency

Quality Metrics

Metric	Target	Measurement
Decision documentation	100%	All decisions have rationale
Progress visibility	Real-time	<5s update latency
Context attribution	100%	All outputs cite sources
User comprehension	>90%	Survey/feedback

Success Output

When this skill is successfully applied, you MUST output:

✅ SKILL COMPLETE: process-transparency

Completed:
- [x] Decision transparency implemented with rationale documentation
- [x] Progress visibility enabled with real-time updates
- [x] Action logging active with full audit trail
- [x] Context transparency established with source attribution

Outputs:
- Decision logs with rationale and evidence
- Progress dashboards with completion metrics
- Action audit trail with timestamps and context
- Context attribution report showing sources

Completion Checklist

Before marking this skill as complete, verify:

All decision points document rationale and evidence
Progress tracking displays real-time status (<5s latency)
All significant actions logged with transparency metadata
Context sources attributed with proper citations
User comprehension validated (survey/feedback if applicable)
Integration points tested (dashboard, audit trail commands)

Failure Indicators

This skill has FAILED if:

❌ Decisions made without documented rationale (black box behavior)
❌ Progress updates delayed or missing (stale information)
❌ Actions taken without logging or visibility
❌ Context sources not cited or attribution incomplete
❌ Users unable to understand agent decision-making
❌ Audit trail incomplete or missing critical events

When NOT to Use

Do NOT use this skill when:

Simple, single-step operations where transparency overhead exceeds value
Internal system operations not visible to end users (use lightweight logging instead)
Real-time performance critical and logging latency unacceptable (consider async logging)
Privacy requirements prevent action logging (use privacy-aware-logging skill instead)
Debugging only (use debugging-transparency skill for development environments)

Use alternatives:

For development debugging: debugging-transparency skill
For lightweight logging: Standard logging without full transparency metadata
For privacy-sensitive operations: privacy-aware-logging skill

Anti-Patterns (Avoid)

Anti-Pattern	Problem	Solution
Silent failures	Users unaware of errors, no debugging info	Always log failures with context and surface to user
Implicit decisions	AI acts as "black box," users can't trust	Document every decision point with rationale
Batch updates only	Progress info becomes stale, poor UX	Implement real-time streaming updates
Technical jargon only	Users can't understand logs	Include human-readable summaries alongside technical details
No context attribution	Can't verify decision validity	Always cite context sources with weights
Over-logging trivial actions	Noise drowns out important signals	Apply progressive disclosure: summary → detail → full audit
Assuming transparency = verbose	Too much info overwhelms users	Use structured logging with expandable sections

Principles

This skill embodies:

#5 Eliminate Ambiguity - Clear visibility into all decisions and actions
#6 Clear, Understandable, Explainable - Human-readable rationale for every choice
#7 Trust and Transparency - Full auditability of agent behavior
#8 No Assumptions - Explicit documentation of context and evidence used

Full Standard: CODITECT-STANDARD-AUTOMATION.md

Commands: /pilot --dashboard, /audit-trail
Skills: task-accountability, audit-logging
Hooks: task-completion, session-handoff
Agents: audit-trail-manager

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

Purpose​

When to Use​

Transparency Dimensions​

1. Decision Transparency​

2. Progress Transparency​

3. Action Transparency​

4. Context Transparency​

Implementation Patterns​

Pattern 1: Progressive Disclosure​

Pattern 2: Structured Logging​

Pattern 3: Decision Trees​

Anti-Patterns to Avoid​

Integration Points​

Quality Metrics​

Success Output​

Completion Checklist​

Failure Indicators​

When NOT to Use​

Anti-Patterns (Avoid)​

Principles​

Related Components​