Unified Agent Index (v1.6.2)

v1.6.2 introduces the Unified Agent Index — a parallel skills trigger index alongside the existing rules index, a new agent map in the project schema replacing has_rules, and a Proactive Trigger Check mechanism ensuring agents scan all trigger tables before performing side-effect actions.

What’s New

1. Skills Trigger Index

The skills index follows the same Two-Tier Progressive Disclosure pattern as rules:

The agent reads the index (~10 lines) at session start and loads individual skills on demand when triggers match.

2. Agent Map in project.md

# v1.6.2+ (new)
agent:
  rules: true
  skills: true

# Legacy (deprecated, backward compat kept)
has_rules: true

The /open workflow checks agent.rules first, falls back to has_rules for backward compatibility.

3. Proactive Trigger Check

Before ANY action that edits files, runs commands, or creates artifacts:

1. Scan workspace rules.md trigger table
2. Scan workspace skills.md trigger table
3. Scan project trigger tables (if loaded)
4. IF match found → read the rule/skill file BEFORE acting

Principle: Check THEN act — never act THEN check.

4. Agent Indices Pre-flight (Layer 3)

All 7 workflows with Step 0 now re-read both rules.md AND skills.md:

• /plan, /docs, /backlog, /push, /release, /retro, /end

This guards against attention decay in long conversations — even if /open loaded indices at session start.

Loading Flow

/open
  ↓
Step 2.5a: Read .agent/rules.md (workspace rules index) — ALWAYS
  → Memorize triggers (do NOT read rule files yet)
  ↓
Step 2.5b: Read .agent/skills.md (workspace skills index) — ALWAYS (v1.6.2+)
  → Memorize skill triggers
  ↓
Step 2.5c: agent.rules / agent.skills / has_rules?
  → YES → Read project indices
  → NO  → Skip (workspace indices still loaded)
  ↓
Coding session: agent about to edit files
  → Proactive Trigger Check → Match found → load matching rule/skill → comply

Defense-in-Depth Update

Layer   What                         Where                           Survives
─────   ─────────────────────────    ─────────────────────────────   ────────
  1     Index instructions           agent-behavior.md §4            ⚠️ Lost after truncation
  2     Safety block in output       /open Step 8 report             ✅ In checkpoint summary
  3     Workflow pre-flight          Step 0: re-read rules + skills  ✅ Fresh from disk
  4     File guard headers           <!-- ⚠️ ... --> in file         ✅ Inline in target file

Layer 3 now includes the Proactive Trigger Check — the “Check THEN act” principle.

Workflow Coverage

Migration

No migration required. Fully backward compatible:

• has_rules: true continues to work (fallback in /open)
• Projects without agent map are unaffected
• /para-audit update will suggest migration from has_rules to the agent map

Added in v1.6.2 (FEAT-53). See also: Rule Layers · Defense-in-Depth · Glossary

→ Learn: Rule Layers Architecture → Learn: Defense-in-Depth → Learn: Glossary & Impact Map → Learn: Para-Kit Skill