Forward-Deployed Engineering

Use this bundle when the work is an embedded technical engagement whose success depends on continuity, not merely a recommendation or a code change. This is a normative operating model synthesized from the role observations in source-index.md and from routed specialist methods; the nine-stage sequence is not an externally standardized methodology.

Lifecycle

Discover → Frame → Hypothesize → Build → Evaluate → Deploy → Adopt → Measure → Generalize

Stage	Required question	Minimum output	Stop condition
Discover	What user workflow and problem are real?	Stakeholder/workflow map and unknowns	No recognizable problem or access to the relevant workflow
Frame	What is in scope, who decides, and what outcome matters?	Engagement charter and assumptions/decisions/risks ledger	Authority, constraints, or outcome cannot be named
Hypothesize	What smallest intervention could change the workflow?	Testable hypothesis and decision rule	No falsifiable hypothesis or unsafe test
Build	What operationally complete slice can be built?	Thin-slice implementation plan and owner	Dependencies or permissions are infeasible
Evaluate	What evidence supports quality, safety, and usefulness?	Evaluation and release decision	Baseline, representative evidence, or risk constraints missing
Deploy	Can it be released, recovered, and verified in the authorized environment?	Readiness, rollout, rollback, and verification record	No authorized access, rollback, or release decision
Adopt	Do intended users activate and use it in the target workflow?	Adoption scorecard and intervention record	Adoption failure is unexplained or ownership/support is absent
Measure	Did the capability change the agreed outcome?	Outcome measurement record	Instrumentation cannot distinguish expected from observed
Generalize	What should happen to the local learning?	Productization record and field-learning record	No evidence or receiving owner for the proposed next step

At every stage, read the current charter, workflow map, and ledger and add evidence rather than re-deriving prior decisions. Maintain one engagement charter, one stakeholder/workflow map, and one assumptions-decisions-risks ledger. Each stage records entry evidence, the artifact produced, the accountable decision maker, unresolved unknowns, and the next handoff. Never silently turn an observation into a requirement, a prototype into a production claim, or a local success into a reusable product capability. For the expected depth and evidence labeling of these artifacts, see the worked example engagement.

Where to enter the lifecycle

Enter at the stage that matches what already exists. Do not restart earlier stages unless the current charter's stop conditions require it.

What you already have	Enter at
A stakeholder request or observed workflow opportunity, no validated problem	Discover
Validated problem and stakeholders, no charter	Frame (establish the charter first)
Charter, workflow map, and ledger; no approved requirement	Hypothesize
Approved requirement; thin slice in progress	Build
Built and tested slice; no release decision	Evaluate
Released within the authorized boundary	Adopt
Adopted and measuring against the decision rule	Measure
Post-launch evidence and a generalization question	Generalize
A well-specified bounded change with no continuity need	Route to neckbeard instead

Before acting at any entry point, review the current charter, workflow map, ledger, and preceding stage-handoff record as entry evidence.

Loading protocol

1. Establish the charter before solution design: problem, users, workflow, outcome, scope, authority, constraints, success measure, and stop conditions.
2. Load lifecycle and artifacts and update the shared ledger after every stage.
3. Treat the stage skills in manifest.yaml as candidates. Apply the route-selection conditions, load one primary specialist, and follow its method rather than copying it into this bundle.
4. Before action in a constrained or sensitive environment, load authority and escalation and route access, security, privacy, irreversible, cost, and external-commitment decisions to their authorized owner.
5. Before calling applied AI or any risky capability production-ready, load agent-evals-and-observability and production-readiness, and require baseline, representative and adversarial evidence, constraints, and a release decision.
6. Treat adoption and measured workflow impact as completion conditions, not postscript communications. Use adoption and measurement.
7. Apply the classification rules in generalization and productization, then close with the productization record. Classify local work as configuration, reusable pattern, product capability, transfer/replacement, or retirement, with evidence and an owner.
8. Treat artifacts as private by default and apply the external-sharing gate before they leave the authorized engagement context.

Load only the primary specialist for the active stage. Add a secondary specialist only for a named blocker, risk, or handoff; do not preload every skill in the manifest. If one specialist fully owns the request, stop routing and hand the task to that specialist instead of running the FDE lifecycle.

Epistemic and communication contract

Label each material statement as one of: source fact, engagement observation, inference, recommendation, decision, or commitment. Use the communication reference for concise status and escalation updates. The discovery brief records the overlap audit and source limitations.

Completion and stop rules

The engagement is complete only when the capability is technically verified, deployed within the authorized boundary, adopted by the intended workflow, measured against an agreed outcome, and its learning has a generalization decision. A prototype, demo, or stakeholder approval alone is not completion.

Stop and preserve the ledger when the problem cannot be articulated, authority or access is missing, evidence fails, adoption remains unexplained or below the decision rule, or the next action exceeds the charter. Escalate rather than guess on security, privacy, irreversible changes, material cost, external commitments, or business authority. Route a well-specified repository bug directly to neckbeard and the relevant specialist instead of invoking this lifecycle.

When not to use

Scenario	Reach for	Why
Well-bounded repository change	neckbeard	Owns intake through verified PR and authorized release for a bounded change
Product investment, portfolio, or lifecycle governance	product-lifecycle	Owns investment and lifecycle governance, not delivery continuity
Ongoing reliability ownership (SLOs, alerts, incidents)	site-reliability-engineering	Standing operational ownership, not an embedded engagement
Internal platform design or operation	platform-engineering	Platform ownership, not customer delivery
One discipline fully owns the task	That specialist directly	FDE stops routing when one specialist owns the request
Advisory analysis ending before implementation and adoption	The relevant architecture or decision specialist	FDE completion requires adoption and outcome continuity

File map

Path	Load when
references/discovery-brief.md	Reviewing the boundary, overlap audit, or evidence basis
references/source-index.md	Checking an externally verifiable role claim or refresh date
references/lifecycle-and-artifacts.md	Starting or handing off any lifecycle stage
references/route-selection.md	Selecting one stage specialist without violating its entry boundary
references/authority-and-escalation.md	Working under access, security, privacy, cost, or authority constraints
references/adoption-and-measurement.md	Evaluating activation, workflow adoption, and measurable impact
references/generalization-and-productization.md	Deciding what field work becomes or does not become reusable
references/communication.md	Writing status, decision, escalation, or handoff communication
references/worked-example-engagement.md	Calibrating expected artifact depth or evidence labeling at any stage
templates/	Creating the charter, workflow map, ledger, stage handoff, engagement status, evaluation and release decision, adoption scorecard, outcome measurement record, productization record, or field learning record
manifest.yaml	Reading machine-readable stages, routes, outputs, and conflicts