Skip to content

[step-names] Align "pre-fetch" step-name spelling and use imperative verbs in workflow-authored steps #75087

Closed as not planned

Description

@github-actions

Step Name Alignment Issues

Scanned 56,905 step records and 760 unique step names across 356 lock files. The compiler-generated steps are consistent. The findings below are in workflow-authored steps (name: in the .md sources, compiled into the .lock.yml files).

1. Medium: "pre-fetch" family is spelled six different ways

The same concept ("run a deterministic step before the agent to gather or prepare data") appears as Pre-fetch, Prefetch, Pre-compute, Precompute, Preinstall, Predownload, Preload, and Pre-flight.

Current step name Workflow source Suggested name
Pre-fetch PR data test-quality-sentinel.md keep (reference form)
Pre-fetch PR diff and review comments (x4) impeccable-skills-reviewer.md, mattpocock-skills-reviewer.md, ponytail-reviewer.md, pr-code-quality-reviewer.md keep
Pre-fetch ADR gate PR context design-decision-gate.md keep
Pre-fetch moderation context ai-moderator.md keep
Pre-fetch release and merged PR data weekly-blog-post-writer.md keep
Prefetch dependabot burner context dependabot-burner.md Pre-fetch Dependabot burner context
Prefetch forecast usage artifacts daily-spending-forecast.md Pre-fetch forecast usage artifacts
Failure investigator prefetch (x2) aw-failure-investigator.md Pre-fetch failure investigator data
Deterministic pre-fetch for failure analysis aw-failure-investigator.md Pre-fetch failure analysis data
Precompute code scanning alert candidates code-scanning-fixer.md Pre-compute code scanning alert candidates
Precompute deterministic findings bot-detection.md Pre-compute deterministic findings
Precompute optimization datasets copilot-centralization-optimizer.md Pre-compute optimization datasets
Precompute schema analysis data schema-consistency-checker.md Pre-compute schema analysis data
Precompute semantic refactor slice semantic-function-refactor.md Pre-compute semantic refactor slice
Pre-compute recent changes governance context uk-ai-operational-resilience.md keep
Predownload models.dev API index daily-model-inventory.md Pre-download models.dev API index
Preload linter source and cache context linter-miner.md Pre-load linter source and cache context
Preinstall Aider CLI (x4), Preinstall Pydantic AI coder agent (x2) daily-code-debt-aider.md, daily-go-test-stubs-aider.md, engine-conformance-aider.md, engine-conformance-pydantic-ai.md, smoke-pydantic.md Pre-install Aider CLI, Pre-install Pydantic AI coder agent
Pre-flight checks, Pre-flight validation (non-fatal), Playwright browser launch preflight unbloat-docs.md, ci-coach.md, (Playwright workflow) Pre-flight everywhere (the last one drops the verb too, see #2)

Suggested convention: hyphenated Pre-<verb> (the majority form already used for Pre-fetch, Pre-compute, Pre-filter, Pre-flight). The Dependabot capitalization fix is incidental.

2. Medium: Several steps are noun phrases instead of imperative verb phrases

About 99% of step names follow the imperative <Verb> <object> pattern. These do not:

Current Workflow source Suggested
Closed PR deduplication guard daily-ambient-context-optimizer.md Guard against closed PR duplicates
PR close-rate metric daily-ambient-context-optimizer.md Compute PR close-rate metric
Token budget guard copilot-opt.md Enforce token budget guard (or Check token budget)
Dependabot burner selection context dependabot-burner.md Select Dependabot burner context
Playwright browser launch preflight Playwright workflow Pre-flight Playwright browser launch
Display Test Config daily-choice-test.md Print test config (Print is the established verb, 1,072 uses; Display is used once)
Load Metrics workflow-health-manager.md Load metrics (capitalization: other steps use sentence case for the object)

3. Low (already tracked, not re-filed): Set up/Check out and Safe Output(s) casing

Still present, and deliberately not duplicated here:

Verification

  • Counts and workflow locations were taken from /tmp/gh-aw/agent/step-alignment-input.json. I did not open each .md source to confirm the exact name: line, other than test-quality-sentinel.md:38 and dependabot-burner.md:76.
  • The workflow-authored .md files need make recompile after renaming.
  • Workflows with a source: frontmatter entry are upstream-managed. Rename those upstream and pull the change in with gh aw update.
  • I could not query existing issues with gh in this run, so I relied on cache memory for the duplicate check.

AI dictionary alignment summary

  • Terms reviewed: pre-fetch/pre-compute family, agent, prompt-context steps, Claude Copilot, Agy.
  • Terms adopted: none new. The existing repository convention (hyphenated Pre- prefix, imperative verbs) is reused.
  • Conflicts resolved: none. Agy (a real engine) and Claude Copilot (Claude served via the Copilot API) are intentional and were not flagged.

Generated by 📋 Step Name Alignment · claude · routed: sonnet55 high · ~26 AIC · ⌖ ~2 AIC · ⊞ 7.1K · ◷

  • expires on Nov 6, 2026, 7:59 PM UTC-08:00

Activity

  1. github-actions commented on Nov 7, 2026

    @github-actions
    ContributorAuthor

    This issue was automatically closed because it expired on 2026-11-07T03:59:35.384Z.

    Closed by Workflow

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions