skills.rest
snewhouse
snewhouseCOMMUNITY
Explore all skills by this author →
𝕏@s_j_newhouseGitHub

aa-ma-execution

Execute AA-MA plans from .claude/dev/active/ with validation and provenance logging.

6|1|Updated Apr 5, 2026
Search Tags:#state management#context injection#agent execution#provenance logging#milestone validation#git commit signing#multi-session automation
One-click install
npx skills add https://github.com/snewhouse/aa-ma-forge --skill aa-ma-execution
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: aa-ma-execution
Source: https://github.com/snewhouse/aa-ma-forge/tree/main/claude-code/skills/aa-ma-execution
Command: npx skills add https://github.com/snewhouse/aa-ma-forge --skill aa-ma-execution

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It prevents context loss and state drift when an agent resumes or continues multi-session work by enforcing structured execution of AA-MA artifacts with validation and provenance logging.

Core Features & Use Cases

  • Execution-mode orchestration (Step/Milestone/Full): Runs the right scope of work with mode-based human-in-the-loop gates and strict validation at milestone boundaries.
  • AA-MA context injection with contamination control: Loads the five artifact files using explicit XML delimiters in a priority order (REFERENCE, TASKS, CONTEXT_LOG, PLAN, PROVENANCE) with token-aware skipping.
  • State synchronization and auditing: Updates tasks, reference, context log, and provenance immediately and blocks progress when required sub-step evidence (non-empty Result Logs) is missing.

Quick Start

Use the aa-ma-execution skill when you see an active AA-MA task directory under .claude/dev/active/ or when you ask to continue the named task.

Frequently Asked Questions about aa-ma-execution

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I resume a multi-session agent workflow without losing context or state drift?▼

Multi-session agent workflows are resumed by executing AA-MA task artifacts from the .claude/dev/active/ directory, which injects structured context using explicit XML delimiters to prevent state drift and context loss. It validates state and logs provenance before continuing.

How does context injection work for long-horizon agent execution?▼

Context injection for long-horizon agent execution loads five artifact files using explicit XML delimiters in a strict priority order: REFERENCE, TASKS, CONTEXT_LOG, PLAN, and PROVENANCE, with token-aware skipping to manage context window limits.

How do I enforce milestone validation and human-in-the-loop gates during automated task execution?▼

Milestone validation and human-in-the-loop gates are enforced by executing work in Step, Milestone, or Full modes, which block progress when required sub-step evidence like non-empty Result Logs is missing at strict milestone boundaries.

Can I run multi-session automation plans that require signed git commits and provenance logging?▼

Multi-session automation plans requiring signed git commits and provenance logging are supported through strict integrity verification, immediate state synchronization, and signed commit finalization requirements built into the execution lifecycle.

What's the best way to prevent state loss when an agent resumes long-horizon work across multiple sessions?▼

State loss during long-horizon work resumption is prevented by immediately updating tasks, reference, context log, and provenance files upon execution, while blocking progress when required sub-step evidence is missing.

Do I need structured task artifacts in .claude/dev/active/ to use multi-session agent execution?▼

Structured task artifacts in .claude/dev/active/ are required, as the execution engine resumes the correct step, milestone, or full workflow by reading these validated task artifacts to ensure proper context injection and state synchronization.

Related Skills

Explore More Software Engineering Skills→
snewhousesnewhouse
6

goal-condition-synthesis

Synthesize falsifiable Claude Code /goal completion conditions from AA-MA plan artifacts.

Community
Advanced
HenquewlHenquewl
7

finishing-a-development-branch

Verify test suites and finalize git development branches through merge or cleanup.

Community
Advanced
nnexainnexai
1

gsd-autonomous

Automate multi-phase project milestones through a discuss-plan-execute loop.

Community
Advanced
MicrotronXMicrotronX
29

mxOrchestrate

Manage persistent mxLore workflow stacks and session state via MCP APIs.

Community
Advanced