Portable skill and companion CLI for manifest-driven agentic CLI orchestration with structured contracts, resumable state, orchestrator-owned verification, and bounded self-healing.
agent-threader/
skill/
SKILL.md # Original monolithic skill (preserved)
SPEC.md # Normative v2 architecture specification
schemas/ # JSON schemas for all contracts
templates/ # TypeScript scaffolding (types, parser, orchestrator)
build/
manifest.json # Skill registry and fragment declarations
compile.mjs # Fragment-resolving compiler (7 IDE targets)
fragments/
common/ # Shared rules (workflow, model selection, portability)
domain/ # Deep domain knowledge (architecture, healing, contracts)
meta/ # Skill system metadata (schemas ref, templates ref)
skills/
agent-threader/
agent-threader.md # Composed skill source with {{include:...}}
src/ # Companion CLI (TypeScript)
compiled/ # Machine-generated IDE-specific outputs
platforms/ # Thin translation layers for Codex, Cursor, Claude, Windsurf
install.sh # One-command setup for all supported tools
The architecture separates concerns cleanly:
- the manifest declares work
- the orchestrator owns scheduling, verification, checkpointing, and healing
- adapters invoke specific CLIs through a uniform interface
- workers and healers emit fenced JSON contracts
- parsers and schemas validate output before state changes
manifest.v2verify_profile.v2task_result.v2heal_decision.v2state.v2
See SPEC.md for the full contract and runtime rules.
bash install.sh # Auto-detect tools and install
bash install.sh --all # Install for all five tools
bash install.sh --cursor # Install for Cursor onlynpm install # Install dependencies
npm run build # Compile TypeScript CLI
npm run compile # Compile skills to all IDE targets
npm run compile:validate # Validate manifest vs source includes
npm run compile:watch # Recompile on changeagent-threader validate-manifest <path> # Validate a manifest.v2 file
agent-threader init-state <manifest-path> # Initialize state from manifest
agent-threader parse-result <log-path> # Extract task_result.v2 from worker log
agent-threader parse-heal <log-path> # Extract heal_decision.v2 from healer log
agent-threader status [state-path] # Display run status
agent-threader logs [state-path] # List log files from state historyAll commands support --json for machine-readable output.
MIT. See LICENSE.