Quick Start¶
This walks through the full mission lifecycle against the deterministic mock adapter: register a project, create a mission, run it, inspect findings and synthesis, resolve a human decision, and read the audit trail. Every command and every output shown below was executed against the real CLI in a throwaway temporary directory before publishing this page.
1. Set up a project to govern¶
Any Git repository can be a managed project. This example uses a fresh one:
2. Initialize Assync in it¶
This scaffolds .assync/ (manifest, rules, boot protocol, handoff
templates) — the marker Assync's CLI looks for, walking upward from the
current directory, before it will run mission commands.
3. Register the project¶
4. Inspect the role pool¶
Prints all 27 built-in executable roles (plus the non-executable Human Governance Authority) with category, risk level, and executability — see Built-in Roles for the full table.
5. Create a mission¶
assync mission create --project-id project_my_project \
--title "Add rate limiting to the API" \
--objective "Add a basic rate limiter to the public API and cover it with tests" \
--required-role lead_programmer --required-role qa_engineer \
--acceptance-criterion "tests pass"
The default risk level is medium, which adds technical_reviewer to
the selected roles automatically — see the workflow section of
Mission Orchestrator. Every selected role
needs an adapter entry in the run config below,
including roles the orchestrator added on its own.
6. Configure adapters and start the mission¶
{
"lead_programmer": {"adapter": "mock", "scenario": "approval"},
"qa_engineer": {"adapter": "mock", "scenario": "blocking_security_finding"},
"technical_reviewer": {"adapter": "mock", "scenario": "non_blocking_finding"}
}
Mission mission_20260728t194346_8e047ff4 -> WAITING_FOR_AUTHORIZATION
Overall recommendation: BLOCK
Decisions requiring human review: ['decision_20260728t194413_038cd757']
The qa_engineer role's mock run reports a blocking high-severity
security finding, so the deterministic synthesis engine recommends
BLOCK and the mission moves to WAITING_FOR_AUTHORIZATION instead of
COMPLETED — see Synthesis Engine.
7. Inspect findings and synthesis¶
Lists both findings from this run: the blocking high/security finding
from qa_engineer and a non-blocking low/quality finding from
technical_reviewer.
Returns the full SynthesisResult: overall_recommendation: BLOCK,
the conflicting role recommendations
(lead_programmer=APPROVE, qa_engineer=BLOCK,
technical_reviewer=APPROVE_WITH_CONDITIONS), the blocking finding, and
recommended_next_authorization: DECISION_GO — synthesis never resolves
this on its own.
8. Resolve the human decision¶
Shows one PENDING decision: "One or more blocking findings remain
open — accept, remediate, or reject before this mission can proceed."
assync decision approve decision_20260728t194413_038cd757 \
--actor you --rationale "Accepted risk; will remediate in a follow-up mission."
Approving a decision does not, by itself, complete the mission
This MVP has no automatic mission-state re-evaluation after a
decision resolves — the mission stays in
WAITING_FOR_AUTHORIZATION until a separate mission-level action is
taken. See Decision Queue — MVP Limitations.
assync mission abort <id> --reason ... --actor ... is the only
other mission-level transition exposed by the CLI today.
9. Read the audit trail¶
Prints the complete, append-only sequence for this mission: mission
created → state transition READY → ACTIVE → each role's run completing
→ each finding created → the decision created → the state transition
ACTIVE → WAITING_FOR_AUTHORIZATION → the decision resolved APPROVED
with its recorded rationale. Nothing in this log is ever edited after the
fact — see the atomicity section of Runtime Storage.
Where to go next¶
- Mission Orchestrator for the full state machine and workflow.
- Authorization Model for how capabilities are actually calculated and enforced.
- Adapter Architecture for the restricted local-command adapter, if you want to run real (not mock) commands.
- CLI Reference for every command and flag.