Feasibility Check
Before commitment, assess the killer risks honestly and put a recommendation bound to its evidence to the accountable owner.
When to use
- Before committing to a project or initiative (the go/no-go moment).
- When feasibility is genuinely uncertain: new technology, hard constraints, unknown data.
- Skip only when an existing feasibility section already fills every
dimension row from evidence and no killer risk's confidence is
unknown. - Solo owner: when the user is the only stakeholder, they own every dimension row and every go-if condition, so Step 1.1's owner lookup is skipped.
Step 1: Assess
Open assets/feasibility-section.md before starting; each step fills its
section.
- Fill every dimension row with its accountable owner. "Technically possible" answers one row, not the others. Every owner this section names, go-if conditions included, comes from the approved goals' decision rule and stakeholder roles. If no owner is named there, ask the user who owns it; never invent one.
- For each killer risk, record the evidence source, its freshness, and the
confidence. Evidence recalled, not read or run in this session, has
confidence
unknown. - Turn the top unknown into a bounded question and record it as a go-if
condition whose evidence is a
prototypingspike. One spike answers one dimension. - Fill Option Zero with evidence for whether not building, an existing tool, configuration, or process, or a smaller initiative meets the goal and guardrails.
Step 2: Recommend and present
Write go / no-go / go-if, bound to the exact version of the recommendation. For each go-if condition, record a stable ID, its gate or its evidence, an owner, an expiry, its state (
pending / satisfied / failed), and the abort response.Write the
## Feasibilitysection, which is never edited once its identity is issued, to.sdlc-skills/briefs/{{YYYY-MM-DD}}-{{topic}}.mdor the user-set path, and leave the approved sections around it untouched.Present:
Feasibility {{path}} — version {{identity (per template)}} Recommendation: {{go | no-go | go-if}} Top risks: {{list with confidence}} Conditions: {{each with owner and gate}} 1. Go 2. Go-if every named condition is met 3. No-go 4. Request changes 5. Cancel Recommendation: {{option}} — {{one sentence of evidence}}.Ask this and end the turn;
clarifying-intentowns what closes it. On option 4, recordchanges requested, write a successor section with the change, and present it again.On a go-if, REQUIRED SUB-SKILL: invoke
prototypingfor a condition whose evidence is a spike. Only the named evidence and owner move a condition's external state (its state in the ledger the template names: add to it, never rewrite it). On expiry, or on any change to a bound input, gate, evidence, owner, or freshness, the condition is invalid and the decision is reopened.REQUIRED SUB-SKILL: on a direct go, or a go-if with every condition
satisfied, invokescopingwhen the boundary is next. Never impose a phase already complete.
Gotchas
- Filling in Option Zero with a bare assertion ("no existing tool fits") instead of the same evidence the other dimensions require still counts as skipped in substance; it only looks done.
- Giving a risk a confidence level it has not earned, rather than recording "unknown", passes off a risk with no real evidence as one that reads as already assessed.