tizen-sdk-skills-review
Skill19 starsChanged 22 days agoApache-2.0View on GitHub
Review a pull request that touches tizen-sdk-skills/ against AGENTS.md, CONTRIBUTING.md and scripts/check-pr.js, then post the review as a pull request comment. Use when asked to review a pull request in this repository, for example "review PR 49" or "PR 49 리뷰해줘".
- Installs packages
In this repository3 other files it gives its agents.
- AGENTS.mdAGENTS.md
- llms.txtllms.txt
- tizen-sdk-skills-reviewSkill
The file
We split this skill into the rules it gives an agent
10 rules. About 0 are common, 7 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Distinctive: line 17
When a rule is ambiguous, use evidence in this order:
- Very specific: line 36
Stop and tell the user if the pull request does not touch tizen-sdk-skills/.
- Distinctive: line 47
If git status --porcelain is not empty, stop and ask the user.
- Very specific: line 64
Every ERROR is blocking. Every WARN is blocking unless the pull request description explains it.
- Distinctive: line 67
Run the checks that match what the pull request changes.
- Distinctive: line 85
Read the whole diff, then the surrounding code of each changed function.
- Distinctive: line 90
Process handling: Every spawned process has its exit code checked, and failures produce a clear error in the Standard JSON Envelope.
- Distinctive: line 103
Publication safety: No internal hostnames, private IP addresses, personal paths or credentials, including inside added images and …
- Distinctive: line 118
Write the comment in English to a file in your scratch directory, using this structure.
- Very specific: line 160
If the agent was not given its model name or ID, write unknown for that part rather than guessing.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Reviewing tizen-sdk-skills pull requests
Review a pull request against the repository's own rules, run the automated check, and post one review comment written in English. A reviewer runs this from their own checkout; the comment is posted from their GitHub account.
Comment only. Never approve, request changes, merge, push, or edit the pull request branch.
Sources of truth
When a rule is ambiguous, use evidence in this order:
AGENTS.mdandtizen-sdk-skills/CONTRIBUTING.mdtizen-sdk-skills/GOVERNANCE.mdsection 2.2 (code review) and the README's Harness Separation Principle- Neighbouring code and documents in the area the pull request touches
Workflow
1. Gather the pull request
gh pr view <N> --json number,title,body,author,baseRefName,headRefName,headRefOid,files,commits
gh pr view <N> --comments
gh pr diff <N>Stop and tell the user if the pull request does not touch tizen-sdk-skills/.
2. Collect prior findings
Earlier review comments start with ## Review:. List every blocking and
non-blocking finding they raised; each one is rechecked against the current
head in step 5. Read the author's replies too: a finding the author explained
is resolved if the explanation holds.
3. Check out the pull request
If git status --porcelain is not empty, stop and ask the user; do not stash
or discard their work. Note the current branch so you can return to it.
gh pr checkout <N>
git fetch origin <base>4. Run the automated checks
cd tizen-sdk-skills
node scripts/check-pr.js --base origin/<base> --format jsonlThe run always ends with check-pr: N ERROR, M WARN (K files, T s). If that
line is missing, the check did not run; fix the --base revision and rerun.
Every ERROR is blocking. Every WARN is blocking unless the pull request
description explains it.
Run the checks that match what the pull request changes. Safe tier only; never run mutating or device-tier tests, and never run anything that needs a Tizen SDK or device.
| Changed | Run |
common/lib/** | node common/lib/tests/run-all.js --filter <module> |
common/hooks/** | bash common/hooks/hooks.test.sh |
any SKILL.md, plugin-cache.js | node scripts/rewrite-runner-snippets.js --check |
| any source file | node scripts/add-spdx-headers.js --check |
tizen-cli/** | cd tizen-cli && pnpm install --frozen-lockfile && pnpm exec tsc --noEmit |
tests/** | cd tests && npm ci && npm run lint |
Record every command you ran and its result for the comment. When a command cannot run locally (missing tool, network), say so instead of guessing.
5. Review what the check cannot judge
Read the whole diff, then the surrounding code of each changed function. Check:
- Harness separation. Logic lives in
common/;claude/,cline/,codex/andgemini/hold thin wrappers and hook adapters only.tizen-cli/andvscode/do not vendor copies ofcommon/. - Process handling. Every spawned process has its exit code checked, and failures produce a clear error in the Standard JSON Envelope.
- Cross-platform parity. A
.shchange has the matching.ps1change (and the reverse); paths are built withpathAPIs, not string joins. - Behaviour change without a test. New or changed behaviour in
common/lib/comes with a*.test.jscase; the Testing section of the pull request records which tiers were run. - Internal-only content. Nothing that depends on Samsung-internal
services appears outside
internal-onlyfences or listed paths (CONTRIBUTING.md section 8). - Documentation. Paired documents change together, and command or skill
counts stay consistent across
README*.md,docs/SKILLS_REFERENCE*.mdandtests/README.md. - Publication safety. No internal hostnames, private IP addresses, personal paths or credentials, including inside added images and screenshots. Look at every image the pull request adds.
- Pull request hygiene. The title follows Conventional Commits with the plugin as scope, the pull request covers one topic, and every template section is filled in.
- Prior findings. Each finding from step 2 is resolved, still open, or superseded; say which and why.
Cite every finding as path:line relative to the repository root, say why it
matters, and give a concrete fix. Do not report style that ESLint or Prettier
already enforce.
6. Draft the comment
Write the comment in English to a file in your scratch directory, using this structure. Omit a section that would be empty, except Automated checks.
## Review: <pull request title>
Reviewed PR #<N> (HEAD `<short sha>`) against `AGENTS.md`,
`tizen-sdk-skills/CONTRIBUTING.md` and `.claude/skills/tizen-sdk-skills-review/SKILL.md`.
### Change summary
<two to four bullets: what changes and where>
### Automated checks
- `node scripts/check-pr.js --base origin/<base>` → **N ERROR, M WARN (K files)** ✅/❌
- <each other command> → <result> ✅/❌
### Prior findings rechecked
**1. <finding> — RESOLVED ✅ / STILL OPEN ❌**
<one or two sentences>
### Blocking findings
**1. <title>** (`path:line`)
<what is wrong, why it matters, and the fix>
### Non-blocking suggestions
- <suggestion> (`path:line`)
### Verdict
<"Ready to merge once the blocking findings are fixed" or "No blocking findings">
---
_Reviewed with <model name> (`<model id>`) in <host>._End every comment with that model line, including re-reviews, so readers can tell which model produced each pass. Fill it from what the running agent was told about itself, for example:
_Reviewed with Opus 5.5 (`claude-opus-5-5`) in Claude Code._If the agent was not given its model name or ID, write unknown for that part
rather than guessing.
7. Confirm, then post
Show the draft to the user and post it only after they confirm:
gh pr comment <N> --body-file <draft-file>Then return to the branch noted in step 3 and report the comment URL.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

