twg

Run TWG for requests needing Atlassian or company context; do not merely recommend it or ask the user to run it. Start from a specific anchor: Jira key or URL, page, person, project, goal, repo, or time window. Answer from read-only command results. If the route is unclear: twg help <terms>, twg help describe <path>, twg help discover-skills "<intent>".

Overview

Load the narrowest companion, then use its TWG route:

• ../twg-jira/SKILL.md for Jira workitems, projects, boards, sprints, and writes.
• ../twg-confluence/SKILL.md for Confluence content, spaces, and authoring.
• ../twg-space-creation/SKILL.md to create or clone Confluence spaces.
• ../twg-status-rollups/SKILL.md for project/goal status, launch/go-no-go readiness, and org/leadership rollups; it precedes ../twg-engineering-work/SKILL.md for PRs.
• ../twg-context-discovery/SKILL.md for dependency maps, repos, and OOO catch-ups.
• ../twg-agentic-search/SKILL.md for deep internal research with Rovo.
• ../twg-responsibility-routing/SKILL.md for owners/SMEs, approvers, escalation.
• ../twg-engineering-work/SKILL.md for code/repo discovery, PRs, and contributors.
• ../twg-jira-resolve-merged-work/SKILL.md for stale Jira work with merged PRs.
• ../twg-operational-health/SKILL.md for incidents/on-call, handoffs, Assets, and risk.
• ../twg-bench-lite/SKILL.md for read-only single-prompt A/B comparisons.
• ../twg-code-review/SKILL.md only when named or asked for additional code-review context.

Invocation And Output

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.

Do not add per-command env prefixes unless requested; hosts may set TWG_AGENT_DEFAULTS=1.

Use stdout_inline first when present. Outside benchmark lanes, inspect output_files.compact only when inline evidence is incomplete; full stdout is the last resort.

In TWG-only benchmark lanes, run only twg. Never use shell utilities or pipelines (jq, rg, date). Use compact/inline output, the prompt's timezone and window, and report gaps. Match the intent to the narrowest companion skill. Let that skill determine the typed route.

Auth/Setup Guard

Do not run setup, login, install, upgrade, upkeep, or credential commands unless explicitly requested for setup/auth/repair. Otherwise report remediation and wait for user direction.

Sandboxed Pipeline Logs

Pipeline logs can redirect to S3. A sandboxed twg bb pipeline get, wait, grep, or tail log request that shows a network-blocked message, S3 hostname, or log-only HTTP 403 while metadata succeeds is a sandbox restriction, not an auth failure. Request an approved unsandboxed retry of that command only, or give the user the exact terminal command. Never request credentials.

Bounded Evidence Loop

1. Classify the anchor: person, team, project, goal, workitem, page, repo, service, or asset.
2. Resolve once; fetch evidence that changes status, risk, decision, or action.
3. Rank candidates, hydrate the set in one batched call, then synthesize.
4. Stop after the first policy denial; stop after the same auth, ACL, contract, or backend error twice.

Batch Reads

One call per entity is the costliest mistake: every call re-submits the whole conversation. When a get accepts a repeated identifier - live help marks these "one or more" - pass the whole set in one call with --agent-fields @compact, about twenty IDs at a time. Otherwise answer from a query or tree route (... query, pr-tree, work-tree, workitem-tree, org-tree, context) rather than hydrating each entity, or hydrate a ranked sample and say what was omitted. Pick the projection before the batch call; re-running one batch to change --agent-fields pays for the set twice. After five same-subcommand calls with different IDs, stop and re-route.

Command Discovery

• Use twg rovo search "<topic>" [--limit <n>] for top-K discovery; explicit --app preflights.
• Trello: twg trello search "<query>"; no workspace scope.
• Run twg rovo list-apps -o json before an explicit Rovo --app; follow the returned auth action.
• Activity history and fuzzy discovery are separate surfaces:
  • twg docs query --since <duration> is user document activity, not title/content search.
  • twg work query defaults to seven days of authored work; other activity requires --activity / --include-viewed.
  • twg docs search "<topic>" is fuzzy document discovery; twg work search "<topic>" is tenant-wide work discovery.
  • Never pass topic text to docs query or work query; use the matching search.
• Resolve URLs, keys, ARIs, and names, then hydrate stable IDs.
• Jira: jira workitem search <text...> for fuzzy text, jira workitem query --jql for structured JQL, rovo search <text...> --app jira for semantic.
• Command shape guardrails:
  • work query uses --scope me|user, never --scope global.
  • Inferred teams (ari:cloud:graph-store::inferred-team/...) need explicit --include-inferred; see references/inferred-teams.md.
• Use search-code; omit --app so all available indexed SCM surfaces are searched; use --repo only as a discovery anchor; widen after generated-doc or incomplete hits.

Assets / CMDB graph

Traversal (object↔owner/team, Jira↔object) → assets graph; see references/ASSETS_GRAPH.md. No hop → assets search, assets query --aql, assets object get.

Load The Narrowest Companion

See Overview.

Rules

• Never guess IDs, flags, slugs, ARIs, or mutation contracts.
• For writes, load the product skill and follow live help.
• Avoid local inspection, caches, or schema probes unless local state is requested.
• For writes, read current state and state the mutation unless execution was requested.