continue-project
/continue-project
Use when a cloned, transferred, or paused project must be resumed, the next safe work is unclear, or shared and local progress need reconstruction.
- Author
- kcrmin ↗
- Repository
- Stackcord ↗
- Licence
- Apache-2.0
- Last repository change
- 12.09.2026
Assessment and security
- Files in this package
- 2
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 12.09.2026, without references to retired models or APIs.
Apache-2.0, otwarta: można używać i zmieniać, z zachowaniem licencji i informacji o zmianach.
What it does and when to use it
Use when a cloned, transferred, or paused project must be resumed, the next safe work is unclear, or shared and local progress need reconstruction.
Original author description
Use when a cloned, transferred, or paused project must be resumed, the next safe work is unclear, or shared and local progress need reconstruction.
How it works
SKILL.md has no level-two headings · 19 lines, 606 words
SKILL.md preview
--- name: continue-project description: Use when a cloned, transferred, or paused project must be resumed, the next safe work is unclear, or shared and local progress need reconstruction. --- # Continue Project Recover from repository evidence, not chat memory. Present one plain-language next action while keeping coordination mechanics in the background. 1. Resolve the CLI from `STACKCORD_CLI`, a repository build, or `PATH`, then run `stackcord status --json` before mutation. Resolve the orchestration root when invoked inside a frontend, backend, or other child workspace. On an explicit continuation request, an absent CLI may trigger an offer to run the matching verified `scripts/bootstrap-cli.sh` or `scripts/bootstrap-cli.ps1` with an explicit release URL, version, and install directory; lifecycle hooks never install it. 2. Re-read actual branch, dirty state, upstream, ahead/behind/diverged state, worktrees, workspace commits, root submodule pointers, approved product meaning, service contracts, external UI authority, the exact UI baseline and source fingerprints, dependent frontend work, evidence, selected live task provider, and `.harness/git-conventions.yaml` when present. Use `$use-git-conventions` before proposing a new branch, commit, pull request, or issue. 3. If GitHub, Jira, Beads, or another external source is selected, use only its installed authenticated connector or CLI. Re-read every changed or dependency-relevant mapped item, normalize the exact revision, and run `stackcord work provider reconcile --apply`. After clone, recover the Git semantic reservation but keep external status unknown until this refresh succeeds. Never present a cached snapshot as live. 4. Summarize confirmed facts, stale derivations, unknown external state, blockers, active ownership, and local-only work. Call out a missing UI checkout, dirty UI workspace, unpublished baseline commit, root-pointer mismatch, or frontend work tied to an older baseline in ordinary language. Explain whether ownership came from the external provider, the recoverable Git reservation, or both. Do not expose IDs or internal storage paths unless diagnosing them is necessary. 5. Never auto-pull, rebase, stash, reset, initialize submodules, move pointers, or claim that unpushed commits are shared. 6. If discovery is unfinished, first run `stackcord project discovery --root <root> --json` (or `--draft <draft-root>` before initialization). Preserve accepted decisions; show current/completed/remaining sections and the approximate remaining-question range. Group eligible independent routine questions and keep sensitive questions explicit as described in [discovery](../../references/discovery.md). Do not run implementation work selection while a discovery blocker needs an answer. Otherwise, when state is coherent, run `stackcord work next --json`. Exclude work blocked by unresolved policy, scenario, contract, migration, UI, dependency, ownership, evidence, or pointer conflicts. 7. Recommend one small reviewable next action using user value, dependency order, unblock impact, and risk. Name the affected workspace, acceptance reference, and first failing test. Ask only when equal safe choices change product intent or Git reconciliation needs human direction. Example: after “이 프로젝트 이어서 해,” report whether frontend, backend, root pointers, and task status are current; then recommend one dependency-ready change instead of restarting discovery. Read [context recovery](../../references/context-recovery.md), [workflow](../../references/workflow.md), [safety](../../references/safety.md), and [tool selection](../../references/tool-selection.md). If the CLI is absent, follow the repo-local fallback and state which checks remain unverified. On first explicit use after the project root exists, run `stackcord setup --json` to read the local UI preference. When no choice is recorded, offer the optional bundled local control center with explicit accept or decline. Accept runs `stackcord setup --ui enable --apply`; decline runs `stackcord setup --ui disable --apply`. Use the resolved root for all commands. Never treat silence as acceptance. Preserve `.harness/local/control-center.json` across Codex and Claude sessions; do not repeat the offer once decided. Later explicit enable/disable uses the same commands, and `stackcord setup --ui ask --apply` resets the preference on request. Launch `stackcord dashboard` only on request. No separate UI download is needed; disabling it preserves all project state and CLI capabilities. Never write durable project state into a plugin cache or host-specific plugin data directory. Both hosts read the same committed `.harness/`, `specs/`, `contracts/`, and local state and use CLI revision checks for writes.
Files in this package (2)
- SKILL.md 5 KB
- agents/openai.yaml 1 KB
Versions
| Repository | Score | Licence |
|---|---|---|
| This page · kcrmin/Stackcord | 96 | Apache-2.0 |
| glifxyz/glif-mcp-server | 82 | MIT |
Database snapshot: 13.11.2026 · package version 4665458b59ba (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 100whats-next
Use when the user asks what's next, where the project stands, for a status update, to prioritise the backlog, or to pick up work where it was left off — an open PR, an unfinished branch or the next ticket. - 96next-task
Use when the user asks what's next, or after a PR merges. - 100session-resume
Cold-start resume — load all memory and project state to continue any interrupted project. Use when: resume, continue, new session, where were we, restart project, pick up, interrupted, cold start.