Skip to content
Back to skills

Continuous Alignment

ASecurity

Autonomous continuous alignment and self-evolving project intelligence engine. Distills session transcripts, enforces a 200-line budget on AGENTS.md, manages path-scoped rules, records living ADRs, and compiles roadmap visualizers via Antigravity hooks. Trigger via /align or /evolve.

  • 2 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 3, 2026
ai-agentspythongobashdatabase

Works with

  • cli
  • mcp

Security analysis

A100/100

Pro scans all 10 files and shows the line behind each finding

Scanned September 20, 2026

npx -y skills add ksprashu/agent-skill-forge --skill continuous-alignment --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Continuous Alignment?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Continuous Alignment
[![Security: A β€” Skills Directory](https://www.skillsdirectory.com/api/skills/ksprashu-continuous-alignment/badge)](https://www.skillsdirectory.com/skills/ksprashu-continuous-alignment)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: continuous-alignment
description: Autonomous continuous alignment and self-evolving project intelligence engine. Distills session transcripts, enforces a 200-line budget on AGENTS.md, manages path-scoped rules, records living ADRs, and compiles roadmap visualizers via Antigravity hooks. Trigger via /align or /evolve.
---

# Continuous Alignment & Project Evolution Engine

Maintain a growing, self-evolving vision, mission, direction, rules, and operational instructions for your software project as development progresses.

---

## 🎯 Goal
Prevent architectural drift and context rot by continuously distilling session insights, enforcing a strict 200-line token budget on root `AGENTS.md`, cascading path-scoped rules to `.agents/rules/*.md`, compiling living Architectural Decision Records (MADRs), and maintaining an interactive SVG roadmap.

---

## πŸ—οΈ 4-Tier Memory Topology

```
β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚ Tier 1: Hot Invariants (`AGENTS.md` / `GEMINI.md`)                     β”‚
β”‚ β€’ Capped at max 200 lines (~1,500 tokens)                             β”‚
β”‚ β€’ Fast-path commands, negative constraints, root gotchas               β”‚
β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€
β”‚ Tier 2: Path-Scoped Rules (`.agents/rules/*.md`)                       β”‚
β”‚ β€’ Subsystem rules progressively loaded via path proximity             β”‚
β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€
β”‚ Tier 3: Strategic Direction (`docs/VISION.md`, `docs/ROADMAP.md`)      β”‚
β”‚ β€’ Living roadmap, milestone tracking, glowing SVG branching visualizer β”‚
β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€
β”‚ Tier 4: Architectural Decision Records (`.gemini/knowledge/ADRs/`)     β”‚
β”‚ β€’ MADR formatted records capturing context, decisions, and trade-offs  β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
```

---

## πŸ“‹ Operational Workflows

### 1. Autonomous Turn Distillation (Antigravity `Stop` Hook)
When an agent turn completes, Antigravity executes `skills/continuous-alignment/scripts/distill_session.py`:
1. Reads the turn execution `transcript.jsonl`.
2. Extracts verified user constraints, error troubleshooting fixes, and CLI commands.
3. Appends deduplicated entries to `.gemini/knowledge/memories.jsonl`.
4. Triggers rule consolidation and living roadmap synchronization.

### 2. Manual Alignment Trigger (`/align` or `/evolve`)
To manually reconcile project instructions, review living ADRs, or prune outdated rules:
```bash
# 1. Distill recent session learnings
python skills/continuous-alignment/scripts/distill_session.py

# 2. Synchronize AGENTS.md and enforce 200-line budget limit
python skills/continuous-alignment/scripts/sync_agents_rules.py

# 3. Compile living ADRs, VISION.md, and ROADMAP.md
python skills/continuous-alignment/scripts/compile_roadmap.py
```

### 3. Context Pulse Inspection (`PreInvocation` Hook)
Inspect the active workspace invariants and recommended commands injected before model turns:
```bash
python skills/continuous-alignment/scripts/sync_agents_rules.py --pulse
```

### 4. Multi-Agent Work Swarm Milestone Distillation
When executing under the `work` skill, each completed milestone emits an adversarial review and auditor `handoff.md`. The continuous alignment engine automatically distills architectural decisions, verified interface contracts, and negative constraints into living MADRs (`.gemini/knowledge/ADRs/`) and path-scoped rules (`.agents/rules/*.md`), ensuring cross-generation coherence across Orchestrator successions.

---

## πŸ›‘οΈ Key Guarantees
- **Zero Third-Party Dependencies**: Pure Python standard library (`json`, `re`, `hashlib`, `tempfile`).
- **Sub-Second Execution**: Hook execution completes in $< 150\text{ms}$.
- **Zero Token Waste**: Strict 200-line budget cap prevents context bloat in root instruction files.
- **Project Tool Isolation**: Scopes external service dependencies (databases, hosting, deployment) under workspace `.agents/plugins/<name>/mcp_config.json`, preventing global configuration clutter and ambient tool pollution.

Files in this skill

  • SKILL.md4.2 KB
  • references/hook_protocol.md2.2 KB
  • references/memory_schema.json2.1 KB
  • scripts/compile_roadmap.py8.7 KB
  • scripts/distill_session.py11.6 KB
  • scripts/sync_agents_rules.py9 KB
  • tests/fixtures/sample_transcript.jsonl463 B
  • tests/test_compile_roadmap.py2.2 KB
  • tests/test_distill.py4.1 KB
  • tests/test_sync_rules.py3.7 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.