agentsclimarketplace
Sign in

Miller text audit

Skill anortham/miller/skills/miller-text-audit

Use when auditing registered Miller workspaces for dangerous strings, deprecated APIs, compatibility markers, secrets-like terms, or other exact text across source, docs, logs, external files, and web imports.From its SKILL.md

Install
npx -y skills add anortham/miller --skill miller-text-audit

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.3 KB, as published. Nobody here has run it

Miller Text Audit

Use this workflow for token-efficient text audits. Prefer content-corpus search over raw grep because it returns ranked snippets, workspace identity, and bounded read coordinates.

Workflow

  1. Check registered workspaces:
miller workspace list
  1. Search each audit term across registered content DBs:
miller content search "dangerous phrase" --workspace-id all --kind source --limit 20 --json
miller content search "deprecated config key" --workspace-id all --kind config --limit 20 --json
miller content search "old docs phrase" --workspace-id all --kind docs --limit 20 --json

Use --kind external_file for imported logs/reports and --kind web for imported web markdown. Use search --mode all-text only when a broad one-workspace union query is more useful than a registered-workspace audit.

  1. Read only bounded windows for evidence:
miller content read --source-id SOURCE_ID --line 120 --context-lines 8

Use the source_id from each content search hit or content list. When the hit came from --workspace-id all, pass its workspace_id back to content read so Miller reads the right registered workspace.

  1. Summarize findings as workspace, file/display path, line, term, and one short snippet. Do not paste full files or full exported JSONL.

Rules

  • Context remains opt-in; do not feed audit hits into context unless the user asks for code context around a finding.
  • Keep exact-term audits exact first. Raise --limit or switch kind only after the targeted query misses.
  • Use content export only for integration feeds, not for interactive audit output.
  • Empty searches and failed reads include compact recovery text and JSON diagnostic_code/next_actions; follow those before falling back to shell search.
  • If --workspace-id all returns no rows, confirm workspaces are registered and content DBs are built before falling back to shell search.

Keep looking

Skills are one crate of 627,801. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.