Skip to content

tmusser/ai-engineering-skills / skills/analyze-mini

analyze-mini

Verified

Risk-triggered consistency checkpoint that detects when task artifacts, assumptions, or verification no longer support a safe build decision.

Analyze Mini

Purpose

Prevent stale or contradictory reasoning from crossing into build-one without turning analysis into a routine gate.

analyze-mini is conditional. The cheap safeguard is an eligibility check over artifacts already being read for build or handoff. Run the full skill only when that check finds a meaningful trigger.

When to use

Run a full analysis when one or more of these are true:

  • HANDOFF.md marks analysis REQUIRED or STALE
  • task-defining inputs changed since the last FRESH analysis: SPEC.md, PLAN.md, TODO.md, scope-freeze notes, selected task, or a decision/assumption that changes implementation
  • the selected task cannot be traced cleanly from acceptance criterion to implementation slice to verification
  • an unresolved decision or assumption could materially change implementation shape
  • compatibility seams, data/metric semantics, or cross-boundary behavior create interpretation risk
  • failed verification caused the implementation strategy or plan to change
  • reconciling a stale handoff changed the live task state

Do not run a full analysis merely because the skill is installed, because a handoff is being written, or because no prior analysis exists.

Missing prior analysis is not, by itself, a trigger.

Staleness is event-driven, not time-driven. A checkpoint does not expire just because time passed.

Inputs

Use the smallest relevant set:

  • Selected task
  • SPEC.md
  • PLAN.md / TODO.md when they exist
  • Scope-freeze boundary
  • Relevant HANDOFF.md state when resuming
  • VERIFY.md when a failure or changed strategy is part of the trigger
  • Only the source files needed to resolve a specific contradiction

Passive eligibility check

Before invoking the full skill, reuse artifacts already loaded for build-one or handoff and ask:

  1. Did a task-defining input change since the last trustworthy checkpoint?
  2. Is there an unresolved decision or assumption that could change the implementation?
  3. Can the selected task still map cleanly from acceptance criterion to verification without contradicting a non-goal, spec ceiling, or scope boundary?
  4. Did a failed attempt or resume reconciliation create a new implementation uncertainty?

If all answers are no, use NOT_NEEDED. Do not perform a broader repo scan just to prove that analysis was skipped.

Checkpoint states

Analysis checkpoint: FRESH | NOT_NEEDED | STALE | REQUIRED

  • FRESH — a full analysis was run against the current task-defining inputs.
  • NOT_NEEDED — the cheap eligibility check found no trigger for a full analysis.
  • STALE — a prior FRESH analysis exists, but a task-defining input changed.
  • REQUIRED — a current trigger exists and a full analysis must run before implementation continues.

FRESH and NOT_NEEDED permit implementation only when no separate blocker exists.

Workflow

  1. Name the trigger. Do not start with a generic repo review.
  2. Identify the smallest artifact delta that could affect the selected task.
  3. Map acceptance criterion -> selected task -> verification evidence.
  4. Check non-goals, spec ceiling, compatibility seams, and scope-freeze boundaries for contradictions.
  5. Check decisions and assumptions that can materially change implementation shape.
  6. For a consequential calculation using mutable data, consider an optional data freeze when retention is appropriate. Preserve the exact extract plus relevant code, parameters, metric definitions and environment references. Record the snapshot location and stamp with calculation evidence; compare a new snapshot when inputs or analysis change. If scripts/aes.py is available, use data freeze|check|diff. A byte stamp alone does not establish data quality or reproducibility.
  7. Separate contradictions from unknowns; do not invent certainty to make the artifacts agree.
  8. Make an implementation readiness call: READY or BLOCKED.
  9. Emit a compact checkpoint:
ANALYSIS CHECKPOINT
State: FRESH
Readiness: READY | BLOCKED
Trigger: ...
Inputs reviewed: ...
Finding: ...
Criterion -> task -> proof: ...
Re-run when: ...
  1. If a handoff is being updated, carry the checkpoint state and reason into HANDOFF.md. Otherwise return the compact checkpoint inline.
  2. Do not create ANALYZE.md for a clean checkpoint. Create or update it only when material findings need durable standalone review or the user explicitly asks for it.

Outputs

  • Compact analysis checkpoint
  • READY or BLOCKED implementation call when full analysis runs
  • Material contradictions, unknowns, or verification gaps
  • Optional ANALYZE.md only when durable standalone findings are justified

Stop conditions

  • NOT_NEEDED is justified by the cheap eligibility check, or
  • the full analysis establishes a current FRESH checkpoint and READY, or
  • a contradiction or unresolved decision makes readiness BLOCKED

Anti-patterns

  • Running analyze-mini before every build-one merely because it is available.
  • Treating a missing prior analysis as analysis debt by default.
  • Re-reading the whole repository to decide whether analysis is needed.
  • Creating ANALYZE.md after every clean consistency check.
  • Expiring a valid checkpoint because of elapsed time rather than a state change.
  • Keeping a checkpoint FRESH after task-defining inputs changed.
  • Starting implementation while a REQUIRED or STALE checkpoint is unresolved.
  • Adding complexity because analysis surfaced an interesting adjacent improvement.
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 194kUpdated Claude CodeCodexVerifiedInstall