Tech Selection — Gated Checklist

A checklist for choosing between technologies, not a scoring rubric. The core insight: these criteria are filters, not sorters — they kill candidates that violate a principle. Keep the candidates and trade-offs visible, then apply the current authorization: choose and verify an authorized reversible implementation; return unresolved business, product or preference trade-offs to the user. Candidate count, duration and confidence language do not determine who decides.

These outcomes end the protocol early:

1. User-owned choice — missing authorization or a genuine unresolved business, product or preference trade-off. Return the evidence and recommendation.
2. Unverified completion claim — withhold the claim, execute authorized verification, and report any remaining acceptance gap.

Not This

• Not for research reports (use deep-research), debugging or bug fixes in already-settled code, or price comparisons between vendors — none of these is choosing a technology.
• Not an interview framework — the user delegates implementation, not direction. Don't ask "which do you prefer" when you can probe and decide.
• Not a scoring rubric — no weighted scores, no "winner" ranking. Filters, then business anchor.
• Not self-certifying — the "why this isn't garbage" defense is written by this skill but must be independently checkable, not self-approved.
• Not a cost gate — budget sets execution tier (which model runs), never whether to do it. Don't use cost as a rejection reason for the user's own projects.
• Not a scope expander — no unrequested features, frameworks, or complexity.
• Not package-size-driven — line count and bundle size are proxies, not quality.

Execution Protocol

Lightweight path. For a small reversible local implementation with no external contract, run Steps 0, 1 and the applicable Step 3 axes. Use the short output below; run the necessary authorized probe before claiming the result works. A JSON library inside a bug fix fits; the project's storage engine or external field contract requires the full protocol even if the change takes under 10 minutes. Duration estimates effort, not authorization. When unsure which path applies, take the full protocol. Both paths obey the stops and any explicit current-task instruction.

Step 0 · Frame

Write these before comparing any candidate:

1. The named business result this choice serves.
2. The named failure mode — what observable phenomenon would prove the choice wrong.

Derive the business result from context, not from the request text. A bare "which DB" carries no result on its face — read the project's decision log, the user's recent corrections, and what this choice unblocks downstream. If no context exists to derive from, say so explicitly rather than fabricating one.

Checkpoint: Cannot derive a business result from any available context → this is an execution task, not a selection task. Exit skill. Business result written as "tests pass" or "pipeline complete" → that's a proxy metric. Rewrite.

Step 1 · Inventory Prior Art

Fixed order: internal/paid assets → external world-class + community solutions → build from scratch (last resort). Tag each candidate with which layer it came from.

Where to look for layer 1: existing credentials and paid-service capability catalogues, installed skills, the current repository's existing pipelines, the project's decision log, and the user's own curated favorites (run favorites-search first when it is installed). Do not limit layer 1 to grep in the current repo — that returns zero hits for paid services and skills, and a zero from a narrow search is not absence.

Layer 2 has a minimum coverage requirement: use a search tool to enumerate what exists, not memory alone. "Searched, found 2" is not coverage — name the search queries run, or state explicitly that no search tool was available and this is a memory-only inventory.

Checkpoint: If the final recommendation falls to layer 3 (build) with no recorded reason from layers 1–2 → flag as 闭门造车. Zero hits from layer 1 must distinguish "searched by structural token" from "searched by remembered name" — the latter's zero hit does not mean absence.

User-named candidates. When the user names a specific option ("use Redis or Postgres?", "should we add a vector store?"), that option enters the candidate set like any other — it is subject to the same axes and verdict. It is neither exempt from filtering (a named candidate is not a requirement) nor disposable (killing a user-named candidate requires naming the axis and the failure mode, exactly like any other). If the user named exactly two options, they are the minimum candidate set; add any layer-1 or layer-2 candidates the inventory surfaced, and say so.

Step 2 · Probe for Evidence

For ingestion, indexing, storage or batch-data choices, load references/batch-data-probes.md and derive the representative cases from the actual consumer contract before comparing engines.

Label evidence by what it establishes: observed execution proves the measured behavior; current official API docs or inspected source can establish a contract or exclude an incompatible candidate, but cannot prove runtime usability or business benefit. README/vendor claims alone leave those claims unknown. Termination clause: max two attempts across methods per candidate; two failures → "this cannot be done now," with the unmeasured claim left unknown.

Checkpoint: Every load-bearing claim names its evidence and scope. Runtime and business-result claims need the corresponding observation; source-contract evidence must not be relabeled as an observed result.

Step 3 · Filter Each Candidate

Read decision axes. For each candidate, give a verdict per axis: pass / fail (name the failure mode) / unknown (needs probe).

A fail on a core axis kills the candidate. A fail on a C-class criterion from references/scoped-criteria.md does not kill — record it as a "declared preference against" note on that candidate and continue. Only core-axis failures remove a candidate from the survivor set.

Checkpoint: unknown is not pass. A candidate carrying unknown does not enter the Step 4 survivor set. Do not output a "winner" from this step.

Step 4 · Triage Survivors — The Core Gate

• 0 survivors: Report which axis killed which candidate, and whether the axis was wrong or the candidate set incomplete. Distinguish unknown from fail; name the stalled probe and recovery condition. Run available authorized probes before asking for information only the user can supply. Never invent a survivor.
• 1 or more survivors: Read references/delegation-contract.md. Check current authorization and consequences, including maintainability and relevant industry practice. For an authorized reversible implementation, choose with evidence, retain alternatives and trade-offs, and execute necessary verification. For an unresolved user-owned choice, return candidates + trade-offs + one recommendation and stop only the dependent action. Do not silently drop rejected candidates.

Checkpoint: Can the output name which candidate was demoted and by which axis? If not, the gate was hollowed out.

Step 5 · Self-Defense Slot

A full-protocol conclusion carries a "why this isn't garbage" paragraph, and it must not be self-certified by this skill alone. The lightweight route carries its named reason and evidence in the short output instead.

Checkpoint: Every sentence in the self-defense traces to evidence or an axis verdict within that evidence's scope. Generic principles (e.g. "it's a mature library") = invalid.

Step 6 · Saturate Irreversible Surfaces

Scan for surfaces that cannot be patched after release: telemetry/events, field and export formats, external contracts, irreversible external actions. If any exist → saturate from v0. This axis is single-scenario evidence — apply it when the choice produces a released artifact, not to internal or revertible changes, and never use it to raise the standard on work that has no irreversible surface.

Checkpoint: Explicitly list irreversible surfaces, or explicitly write "none." Silence = not checked. Revertible local changes do not trigger this step.

Step 7 · Completion Declaration

Distinguish "I verified" from "I claim." Every done statement is followed by what was actually executed and observed.

Checkpoint: Run the authorized checks the agent can perform. Label human acceptance separately; unobserved acceptance stays unconfirmed and must not become a completion claim or block already-authorized verification and delivery.

Output Shape

Lightweight output: business result + falsifier; prior-art layer; chosen local implementation + reason and authorization; necessary probe result or unknown. Keep relevant alternatives visible without requiring the full matrix or unused Steps 4–6 fields.

Full-protocol output: produce the fields below. Include actual verification and human-acceptance status with the decision branch.

1. Business result + failure mode (Step 0).
2. Candidate table (Steps 1–3) — each candidate with its layer tag, its per-axis verdict (pass/fail/unknown), and the probe name behind each pass, or authoritative source evidence for a source-contract verdict. A candidate carrying unknown is marked, not hidden.
3. Survivor demotion record (Step 4) — for each demoted candidate, which axis killed it and the named failure mode. If no candidate was demoted, say "no candidate eliminated" and list the axes that passed everything.
4. Decision branch (Step 4) — declare exactly one of:
   • authorized implementation with chosen survivor, alternatives, trade-offs, authorization and the Step 5 self-defense
   • user-owned choice → STOP with candidates + trade-offs + one recommendation, the missing decision/authorization and the dependent action withheld
   • 0 survivors with the diagnosis (axis wrong vs candidate set incomplete)
5. Self-defense (Step 5) — each sentence traces to a probe or an axis verdict. Write for the chosen implementation or recommendation; for zero survivors, carry the diagnosis instead of defending an invented winner.
6. Irreversible surface list (Step 6) — the surfaces found, or the word "none." Never omit.

Stops

Stop 1 · Multi-candidate human tradeoff

When current authorization leaves a business, product or preference choice to the user, the output is:

• Each surviving candidate
• Its trade-offs (what it costs, what it gives up)
• One recommendation with reasoning

Do not execute that dependent choice before the user resolves it. Keep rejected candidates and their reasons visible; do not fabricate a ranking. Multiple survivors alone do not reopen authorization for a local reversible implementation. New scope, external commitments and irreversible actions require their actual authorization even when only one candidate survives. Reuse prior answers unless later facts or instructions change their applicability.

Stop 2 · Unverified completion claim

The agent executes authorized probes and observes the actual consumer/result; tests passing cannot substitute for that result. If the artifact requires human acceptance, present the concrete result and evidence for acceptance and mark that status unconfirmed until observed. Withhold unsupported completion claims, continue independent authorized work, and identify the exact missing condition when blocked.

Agent Orchestration

Before delegating, read the orchestration contract.

References

Boundary Quick Reference

Before resolving these apparent conflicts, read resolved scope boundaries.