twg-engineering-work
Skill7 starsChanged 2 months agoMITView on GitHub
Use with root `twg` for code search, repositories using an API/package, implementation and reverse-dependency discovery, PR status and reviews, repo contributors, hot areas, and issue-to-PR lookups.
In this repository29 other files it gives its agents.
- AGENTS.mdAGENTS.md
- address-pr-commentsSkill
- build-and-scan-container-imageSkill
- execute-database-migrationSkill
- generate-adrSkill
- generate-service-documentationSkill
- get-pr-commentsSkill
- incident-triageSkill
- install-dev-toolsSkill
- prepare-releaseSkill
- remediate-dependency-vulnerabilitiesSkill
- review-configurationSkill
- twg-agentic-searchSkill
- twg-artifactsSkill
- twg-code-reviewSkill
- twg-confluenceSkill
- twg-context-discoverySkill
- twg-jira-resolve-merged-workSkill
- twg-jiraSkill
- twg-operational-healthSkill
- twg-responsibility-routingSkill
- twgSkill
- twg-space-creationSkill
- twg-status-rollupsSkill
- validate-and-deploy-iacSkill
- validate-and-deploy-kubernetesSkill
- validate-documentationSkill
- validate-github-actionsSkill
- verify-service-healthSkill
The file
We split this skill into the rules it gives an agent
11 rules. About 0 are common, 7 are distinctive, and 4 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.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 11
Use with root twg. Follow documented command shapes.
- Very specific: line 16
Run twg <command>. On shell command not found, use $HOME/.local/bin/twg (macOS/Linux) / $env:LOCALAPPDATA\Programs\twg\bin\twg.exe …
- Very specific: line 52
Use ../twg-code-review/SKILL.md only when named or code review requests additional context.
- Very specific: line 82
Only approval/comment timestamps need per-PR bitbucket pull-requests activity.
- Distinctive: line 98
Keep Bitbucket, GitHub, and Atlassian auth failures separate.
- Distinctive: line 99
If PR graph or repo-wide queries repeatedly fail, make one narrower fallback from known repos, people, workitems, or search anchors, then …
- Distinctive: line 119
Use workitem context for linked PRs, commits, branches, and repos.
- Distinctive: line 141
Do not guess Bitbucket workspace or repo.
- Distinctive: line 142
Do not call a single-entity command per PR or per pipeline when a batch or query route covers the set.
- Distinctive: line 144
Do not fetch every PR body, diff, or comment in a large queue.
- Distinctive: line 145
Do not treat PR counts as impact.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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
twg-engineering-work
Use with root twg. Follow documented command shapes; use focused live twg help
only for missing or uncertain commands, arguments, or output contracts.
CLI launcher fallback
Run twg <command>. On shell command not found, use $HOME/.local/bin/twg
(macOS/Linux) / $env:LOCALAPPDATA\Programs\twg\bin\twg.exe (PowerShell), then
tell user to add that directory to PATH. Do not treat auth or command errors as
PATH failures.
Use When
- "Which PRs are waiting for my review?"
- "Where is this API implemented or used?"
- "Which repositories depend on this package?"
- "Latest PRs for this issue"
- "Who contributed most to this repo/topic?"
- "Repos I created PRs in"
- "Stale reviews"
- "Review flow or bottlenecks"
- "PR-only status for a user, team, or repo", "this person's PRs", "my PRs this week"
- "Open bugs/tasks with PRs in flight"
First Move
Resolve the engineering anchor:
- Repo: identify workspace and repo from URL, local checkout, or repo query.
- PR: resolve exact PR URL, ID, workspace, and repo.
- Workitem: fetch/context the Jira workitem for linked PRs, commits, branches, and repos.
- Topic: resolve/search once, then find linked repos, PRs, and workitems.
- Code: resolve the package, API, symbol, or repository anchor, then load
references/code-search.md.
Provider-native PR commands apply only to their host; Bitbucket activity/comment/task commands never apply to GitHub PRs.
Route Selection
- Use
../twg-code-review/SKILL.mdonly when named or code review requests additional context. Exclude ordinary reviews during testing. - For issue-to-PR lookup, use workitem context before broad PR text search.
- For repo contributors and hot areas, combine PR/commit/file-area signals with ownership and review evidence.
- For PR leadership/team/org rollups, use
twg-status-rollups; this skill adds PR detail. - For person/repo status, collect merged/open PRs for the people, repos, and window.
- Code: load
references/code-search.md. - PRs: load
references/pull-requests.md. - Pipelines and builds: load
references/pipelines.md. - For person-scoped summaries beyond PRs (Jira, docs, meetings, planning,
notifications), use
twg-status-rollupsplus../twg-status-rollups/references/personal-work-summary.md.
Batch The PR Set
Hydrate a PR set in one call, never one call per PR.
- Reviewers, approvals, age, merge state: one
pull-requests get <pr-url…>covers the whole set. Use PR URLs (urlin query results); ARIs giveapproved: null. - Candidates:
bitbucket pull-requests query(author, reviewer, state, window) orpull-requests query(user/org). pr-treereviewer counts mean "listed as reviewer", not approved.- Pipelines:
bitbucket pipeline querylists recent runs only (--limitdefault 15); raise it and check returned timestamps before claiming window coverage.pipeline getonly for failures whose logs matter.commits/deploymentsbatch by ARI. - Only approval/comment timestamps need per-PR
bitbucket pull-requests activity; past ~10 PRs, sample and state it. With no--typeit returns approval, update, and comment events together, so never call it again per event type.
Evidence Policy
- Hydrate PR comments, tasks, pipeline status, and diff only for stale, blocked, central, or high-impact PRs.
- For PR rollups, stop once themes, repos/services, owners, and recency are identified; more PRs of the same theme add nothing.
- For review status, include age, requested reviewers, comments/tasks, approval state, CI/pipeline state, and last activity where exposed.
- For repo/team reports, group by repo, service, or workstream, not person counts.
- Infer themes from PR titles, descriptions, and linked issues, not PR counts.
- Keep Bitbucket, GitHub, and Atlassian auth failures separate.
- If PR graph or repo-wide queries repeatedly fail, make one narrower fallback from known repos, people, workitems, or search anchors, then answer from what succeeded and call out the PR coverage gap.
Recipe Cards
Review Queue
Query reviewer-scoped open PRs; sort by waiting time, requested action, unresolved tasks/comments, failing CI, and relevance. Hydrate only PRs needing action.
Stale Reviews / Review Bottlenecks
Find PRs open or waiting beyond the threshold; group by repo, author, reviewer, and stage. Bottleneck patterns: missing reviewer, unresolved tasks, failing CI, repeated request-changes, owner unavailable.
Issue PRs
Use workitem context for linked PRs, commits, branches, and repos. Fetch PRs only when the user asks for details, status, or next action.
Repo Contributors / Hot Areas
Query PRs/commits for the repo and window; group by files/areas, authors, reviewers, and themes. Hot areas rank changed area, frequency, and ownership.
PR-Based Status Rollup
Resolve org/team first, then collect PRs for members or repos in the window.
Group into themes and repos/services, and call out gaps where PR-only evidence
omits Jira, docs, planning, or customer context. For one person with a broader
prompt, use twg-status-rollups.
Output Shape
Queues: PR, repo, owner, state, reason, next action, evidence. Reports: workstreams, contributors, bottlenecks, risks, gaps, and stable artifact URLs/IDs.
Anti-Patterns
- Do not guess Bitbucket workspace or repo.
- Do not call a single-entity command per PR or per pipeline when a batch or query route covers the set.
- Do not fetch every PR body, diff, or comment in a large queue.
- Do not treat PR counts as impact.
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.

