القائمة
Guru Team Trellis finish entry. Use after task work is committed and Branch Review Gate passed; archives task, records journal, then publishes a non-draft PR.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
Guru Team Trellis continue entry. Use to advance the current task through planning, implementation, commit, and Branch Review Gate before finish-work.
Guru Team Trellis fallback orientation entry. Use when automatic startup context is unavailable, hooks did not run, or the user explicitly asks to reload full Trellis context.
Discovers and injects project-specific coding guidelines from .trellis/spec/ before implementation begins. Reads spec indexes, pre-development checklists, and shared thinking guides for the target package. Use when starting a new coding task, before writing any code, switching to a different package, or needing to refresh project conventions and standards.
Guides collaborative requirements discovery before implementation. Creates task directory, seeds PRD, asks high-value questions one at a time, researches technical choices, and converges on MVP scope. Use when requirements are unclear, there are multiple valid approaches, or the user describes a new feature or complex task.
Comprehensive quality verification: spec compliance, lint, type-check, tests, cross-layer data flow, code reuse, and consistency checks. Use when code is written and needs quality verification, before committing changes, or to catch context drift during long sessions.
Understand and customize the local Trellis architecture inside a user project. Use when modifying .trellis plus platform hooks, settings, agents, skills, commands, prompts, workflows, the channel runtime (trellis channel), bundled runtime agents under .trellis/agents/, selectable workflow templates, registry-backed spec refresh, cross-session memory (trellis mem) generated by trellis init, or AI-facing bundled skills (trellis-channel, trellis-session-insight, trellis-spec-bootstrap) and bundled-skill auto-dispatch flow.
| name | trellis-finish-work |
| description | Guru Team Trellis finish entry. Use after task work is committed and Branch Review Gate passed; archives task, records journal, then publishes a non-draft PR. |
Finish-work is the only user-facing closeout entry. Do not ask the user to remember a separate publish command.
Before reading or writing task-local closeout artifacts, treat task-start-context.json as portable identifiers only (workspace_slug, task_workspace_id, and repo-relative task_artifact_dir). Resolve the machine-local task worktree from the current checkout, .trellis/.runtime/guru-team/**, and git worktree list, then require .trellis/guru-team/scripts/bash/check-workspace-boundary.sh --json --task <task-path> to confirm it. Never read an absolute workspace_path from committed task context.
Before running the helper, generate or review the PR body for GitHub reviewers
who do not know the Trellis task. The body must be Chinese and self-explanatory:
变更摘要, 影响范围, 验证结果, Review Gate, Issue 关闭范围, and
安全说明 must contain concrete content. It must also include a Docs SSOT /
文档同步 result with the plan strategy, durable docs updates or no-update
reason, task delta merged back, task-history-only content, and any follow-up or
current PR limitation. Do not use low-information summaries such as
当前 Trellis task, 已提交实现与文档更新, or 详见 artifact.
Write the reviewed Markdown body to the task-local file
{TASK_DIR}/pr-body.md and pass it with --body-file "{TASK_DIR}/pr-body.md".
A task-local JSON readiness artifact passed with --body-artifact <path> is
also accepted, but the main flow should use the reviewed Markdown file.
Non-draft publish requires one of these reviewed sources; script-generated
generated bodies are preview/draft-only and do not count as readiness
evidence. The readiness/body files are Trellis task metadata and publish reads
the final body from the archived task artifact after archive.
Run the internal Guru Team finish helper as a dry-run readiness preview first:
.trellis/guru-team/scripts/bash/finish-work.sh --json --from-trellis-finish-work \
--body-file "{TASK_DIR}/pr-body.md" \
--dry-run
Only after reviewing the dry-run output, run the formal finish:
.trellis/guru-team/scripts/bash/finish-work.sh --json --from-trellis-finish-work \
--body-file "{TASK_DIR}/pr-body.md"
After dry-run, run resolve-human-artifacts.sh against the active task and
include an active-task Markdown 产物 review 表:
.trellis/guru-team/scripts/bash/resolve-human-artifacts.sh --json --task <task-path>
After the formal finish archives the task, run the resolver again against the
archived task name or archive path and include the archive-path
Markdown 产物 review 表 in the final reply. The table lists only prd.md,
design.md, implement.md, review.md, and pr-body.md; missing files must
not be rendered as Markdown links, and JSON artifacts stay out of the standard
table.
The --from-trellis-finish-work marker is required proof that this explicit finish entrypoint was invoked; do not add it to trellis-continue. The helper verifies the passed Branch Review Gate, allowing only Trellis metadata such as review.md, reviews/*.md, agent-assignment.json, review-gate.json, and PR readiness files after the reviewed HEAD; rejects uncommitted non-metadata changes; archives the active task; records the session journal; commits any remaining Trellis metadata-only changes; then internally pushes the reviewed content branch, runs deterministic remote marketplace init/preview/switch/preset-reapply verification, records schema-valid task-local marketplace-verification.json, replaces the required structured pending evidence in primary/close Issue Scope Ledger entries with real verifier facts, commits exactly the artifact plus ledger as the metadata-only tail, pushes it, cross-validates artifact SHA-256/content HEAD/remote HEAD against the ledger, rechecks the exact two-path tail and Branch Review Gate, and only then creates the PR. Missing, pending, failed, tampered, or stale verification blocks before gh pr create; no tag is created. Durable docs, .trellis/spec/, source, tests, schema, config, scripts, preset, overlay, CI/CD, deployment, migration, or Makefile drift after the gate must return to Phase 2/3; finish-work/archive must not first merge durable docs or patch missing Docs SSOT work. Direct publish-pr is not the normal path and is reserved for explicit recovery/debug after finish-work. The helper does not perform review itself; the gate must already record task-local review.md as review_report digest evidence and raw review_reports[] digest evidence for every review round, and those Markdown reports must already be Chinese human-readable task artifacts except for literal command/path/JSON/HEAD/API/code tokens; when sub-agents were used it should already record agent-assignment.json digest/roles evidence. Docs SSOT reconciliation and any required Middle-platform Knowledge Gate evidence must also be present.