factory-mission

What this skill does

factory-mission is the execution-harness orchestrator of the software factory. It takes a feature description, structures it into a Mission Brief (goal, constraints, non-goals, success criteria), generates an ordered step list, and executes those steps via sequential subagent runs.

It implements the following key factory platform capabilities:

Universal Skill Routing: Decoupled step dispatching. It scans installed skills and hands the inventory to subagents (the model picks which tool fits the step).
Tracker-Agnostic Integration (references/tracker-integration.md): When invoked with --issue <ref>, it pulls ticket context, respects automation/dispatch labels (autonomous/supervised), and writes back status comments + iteration logs.
Inter-Agent Comment Bus (references/tracker-integration.md §Inter-Agent Comment Bus): When tracker-integrated, each step's terminal output (decisions, findings, artifact references — never drafts) is published as a structured comment on the PR/MR/issue. The next step reads previous markers before starting. This is the durable inter-agent memory that survives session boundaries, runtime switches, and pod crashes. Drafts stay on local disk.
Self-Contained Worker Brief: The Mission Brief is persisted to .adlc/workflows/runs/<run_id>/brief.md as a draft. Resumed runs and cross-runtime workers read it from disk — no session-context dependency.
Worktree Isolation: Each run gets its own git worktree. Never touches the user's main checkout. Cleaned up on exit (retained if unsaved work).
Lease-Based Liveness: The state file carries a renewable lease with heartbeat + TTL. Resume can distinguish live, stale, and completed runs.
Stall Detection: After dispatching a subagent, observable progress is checked at a configurable window (default 20 min). A hung agent that passes the circuit breaker is detected and killed.
Lane-Based Dispatch (references/lanes.md): Steps can run on different lanes — inline (this session), agent (fresh session of same CLI for maker/checker separation), or cli:<runtime> (optional cross-vendor). The agent lane is the default for unattended stages.
Scratchpad Tools: Subagents share named, run-private scratchpads (.adlc/workflows/runs/<run_id>/scratchpads/<name>.txt) to compile notes, drafts, and reviews incrementally before publishing.
Workflow Memory & Self-Improvement: Persistent JSONL database (.adlc/workflows/memory.jsonl, workspace-global) stores learnings across runs. factory-learn periodically runs retrospectives to prune/weight memories.
Hierarchical Context Parameters: Workflows and agents reference parameters as {{params.<key>}}, resolved from most specific to least specific: agent < workflow < repository < project < default.
Decoupled Test/Code Separation: In autonomous or supervised modes, it splits implement into sequential test (Test Agent writes failing tests under read-only src/) and code (Implement Agent writes code under read-only tests/) runs — each closed by a mandatory mechanical gate (see Phase 5): the RED gate proves the new suite fails before coding starts; the GREEN gate proves it passes before converge is reached.