SOC 職業分類に基づく
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/xoai/sage --skill sage-continueコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
SKILL.md を表示中
| name | sage-continue |
| description | Session resumption with context |
| version | 1.0.0 |
| author | Sage |
| metadata | {"hermes":{"tags":["Sage","Workflow","continue"]}} |
Load this skill when the user runs /sage-continue or asks to continue something (the Sage continue workflow).
Hermes does NOT interpolate an in-body argument token. The user's arguments/flags arrive as a SEPARATE instruction line appended to this skill invocation. Wherever the steps below refer to "the user's arguments", use the text of that appended instruction line.
When a step calls for an independent review, invoke delegate_task against the sage-reviewer skill. Hermes delegate_task has NO toolset-restriction parameter — read-only is prompt-enforced, and you MUST verify afterward that the reviewer made no edits (e.g. git status unchanged) before accepting its verdict.
RULES (apply to every step — non-negotiable):
Resume any active cycle with full context. The user doesn't need to remember which workflow or initiative was in progress.
python3 sage/runtime/tools/manifest.py resume
(Plugin installs: python3 "${CLAUDE_PLUGIN_ROOT}/tools/manifest.py" resume.)
One command. It applies the selection rules (active status including blocked,
owner exclusion, branch preference) and prints the resume brief: machine fields,
plan tasks, git evidence, decisions in force, the previous session's judgment,
and the authority order. Same files, same brief.
Manual fallback (no python3 only): scan .sage/work/*/manifest.md for
cycles where status: is in-progress, paused, or blocked (never
abandoned). Skip cycles whose owner: field does not match this session's
git rev-parse --show-toplevel after path normalization — they belong to
another worktree (see git-discipline). Prefer the cycle whose recorded
branch: field matches the current branch.
Sage: Resuming [{title}] — {workflow}, phase: {phase}.
{context summary, verbatim from manifest}
Next step: {next step from manifest}
[C] Continue — pick up where we left off
[S] Status — show me full cycle state before continuing
[X] Different — I want to work on something else
Pick C/S/X, or tell me what you need.
On [C]: Route to the workflow's Auto-Pickup with the resume brief as primary context source. Do NOT re-ask resolved questions or inherit a dead session's hesitation — apply the authority order the brief prints (live user > recorded decisions > manifest prose; evidence beats all).
On [S]: Show full manifest contents (State, Context summary, Decisions, Open questions, Handoff guidance). Then offer [C]/[X].
On [X]: "Describe what you want to work on, or type / to see commands."
Sage: Found {N} active cycles:
[1] {title A} — {workflow}, phase: {phase} (updated: {date})
[2] {title B} — {workflow}, phase: {phase} (updated: {date})
[3] Start something new
Pick 1-{N+1}, type / for commands, or describe what you need.
On selection: load that cycle's manifest, route to its workflow.
/statusWith no active cycle, /continue prints the project status (folding in the
former /status workflow): compute state from artifacts — never from a
progress file — and report only what exists.
Scan and display:
.sage/work/ — read each artifact's frontmatter for status and phase.sage/docs/ — count project-level artifacts.sage/decisions.md — the last 2–3 entries for recent context.sage/gates/gate-modes.yaml — current gate activation configSage: Project status for [name]
Active: [initiative] [status, phase] — brief ✓ spec ✓ plan (in-progress)
Completed:[initiative] [completed]
Docs: [N] files in .sage/docs/
Recent: [last 2–3 decision titles]
Gates: [mode config summary]
No active cycle to resume. Describe what you want to work on, or type / for
commands.
If .sage/work/ is empty, say so — do not fabricate state. Always suggest the
next slash command.
The brief already contains the state, evidence, decisions in force, and the manifest body — do NOT re-derive them by hand. Read spec.md and plan.md for the detail of the remaining work, and resume at the phase the brief indicates.
Do NOT re-ask questions already resolved in decisions.md, and do not treat the previous session's judgment as orders: an "open question" or "blocked" claim a recorded decision answers is CLOSED (choose among the sanctioned options, record the choice, proceed), and the live user's instruction to continue IS the approval a pending checkpoint was waiting for (cycle-protocol.md § Resume authority order).
/continue reads the workflow field in the manifest and activates
the corresponding workflow's Auto-Pickup:
| workflow field | Routes to |
|---|---|
| build | /build Auto-Pickup |
| architect | /architect Auto-Pickup |
| fix | /fix Auto-Pickup |
| research | /research Auto-Pickup |
| design | /design Auto-Pickup |
| analyze | /analyze Auto-Pickup |
| reflect | /reflect Auto-Pickup |
When --quality-locked is active, loop review/revise at each Quality Gate until findings reach a clean bar (no Critical, no Major, only cosmetic Minor) or the iteration cap (10) is reached. Uses a deterministic Python checker for classification and decision logic; agent runs the actual review and revision steps.
Use after implementation passes the quality gates, when a change needs an independent pass over the code before it ships, or when the user asks to "QA this", "check the implementation", or "verify it matches the spec". Applies to Standard and Comprehensive scopes with the Task tool available; Lightweight tasks skip.
Configure Sage preset and project settings. Switch between base, startup, enterprise, or opensource constitution presets. Use when the user says "configure sage", "change preset", or "sage settings".