Use when reviewing agent context (SKILL.md, plugin, rule, AGENTS.md) with dora before reporting done. Also when ask-dora routes a Dora job (review, fix, unused, scan), or the user asks to review, diagnose, or clean up Claude, Cursor, Codex, Copilot, or Grok context.
68
85%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Low
Low-risk findings worth noting
You run dora. Assess the agent setup in this project and show what to fix. Review is the gate. Done means exit 0. Ship a repeating job as a pocket agent.
If dora is not on PATH, use npx @hacksmith/doraval.
Review. dora review --quick on a new repo. After an edit, dora review --quick <path>.
<path> is the Skill directory, a plugin root, or ..
More than 10 artifacts: pass a path or --all.
Done: Findings read, exit code read. exit 0 is the only report-done.
Author only when the user asked you to write a Skill, Rule, agent, or plugin.
dora skill new / dora rule new / dora agent new / dora plugin new.
Done: the files exist.
Mechanical fix.
dora fix <path> --dry-run
dora fix <path> --yes--yes applies frontmatter, formatting, and missing fields.
Done: dry-run is clean, or --yes applied.
Judgment. Run Brief. Then replace.
dora fix <path> --brief (or --json).
Each item: message, severity, hint, optional code / docUrl.
Target is that Authored SKILL.md.
Replace one item at a time:
references/.Done: every item replaced or shown. Then step 5.
Re-review. dora review --quick <path> after every edit.
Done: exit 0. That is the only report-done.
Delegated Judge only after step 5 is exit 0 and a later Review (no --quick) is delegated.
Read JUDGE THIS. Evaluate. Fix Findings.
Done: mechanical tiers clean and the rubric applied.
Branch on JSON: output.md.
Pass --yes or --dry-run on fix, conflicts, memory promote, skill remove, skill restore, probe, harness new --accept, harness apply, harness rm, and config set of a secret.
--brief is not a write. A missing write flag is exit 2.
Verbs: commands.md. Exit codes and JSON: output.md.
dora wire --for claude --run records the request the model received. It is not a Finding. Logs stay in ~/.doraval/wire/.
Recurring use case → dora harness new (starts ask-dora / grilling-for-routine). Then dora harness apply <slug>. dora harness boot <slug> is the apply alias. Then hermes mcp login scalekit. Runtime watch after apply, boot, list, pause, and resume. Also dora harness show <slug>, dora harness models, dora harness logs <slug>, dora harness rm <slug> --yes, dora harness pause, dora harness resume, dora harness list, dora harness open.
MUST report done only on exit 0. MUST NOT skip Brief for Judgment items.
saif-shines/doraval
f1834a4
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.