About dstack-core
dstack-core is a support skill for dstack workflows, defining shared contracts for review, feature lifecycle, and delivery checks. It helps workflow authors apply consistent state and mutation rules.
Key features
- 01Review state and finding contracts
- 02Feature resolution through Beads
- 03Standalone issue execution
- 04Lifecycle version checks
- 05Guarded Beads publication
Use cases
- 01Resume a review with durable reviewer state
- 02Select the next ready feature epic
- 03Execute one standalone issue
- 04Check delivery against the base-branch commit
Install
Copy the command and run it in a terminal.
npx skills add https:/After installing, type/dstack-corein Claude Code to confirm it loaded.
Bundled files
References and scripts installed along with SKILL.md.
assets/pi-reviewers/dstack-clarity-reviewer.mddstack-delivery-integrity-reviewer.mddstack-implementation-reviewer.mddstack-readiness-reviewer.mddstack-task-reviewer.mdreferences/INTERACTION-BOUNDARY.mdPI-REVIEWER-ROSTER.mdREVIEW-FINDINGS.mdREVIEW-STATE.mdSKILL-VERSION.mdTRUST-AND-AUTHORITY.mdscripts/append-review-note.pybeads_hooks.pybeads_workflow_lock.pybeads-workflow-lock.pybuild-beads-review-projection.pycheck-skill-version.pyfinalize-feature-delivery.pyguarded-beads-push.pymigrate-review-topology.pypromote-legacy-feature.pyreconcile-beads-interactions.pyresolve-feature.pyreview-state.pyrun-workflow.pysync-pi-reviewers.pyverify-delivery-state.pyDetails
From the project READMEdstack Core Contracts
This support skill contains the shared contracts used by the dstack workflows.
Review orchestration uses references/REVIEW-STATE.md for durable reviewer identity,
source-boundary assignments, resumption, and replacement evidence,
references/REVIEW-FINDINGS.md for current finding dispositions, and the optional
references/PI-REVIEWER-ROSTER.md adapter for mapping logical roles to named Pi
reviewer definitions without changing the tool-agnostic contract. Use scripts/sync-pi-reviewers.py only through its
explicit user-selected target; normal skill installation never mutates Pi agent directories.
build-beads-review-projection.py supplies execution readiness with a validated transient graph without reading Beads
storage directly, and append-review-note.py validates and byte-preserves one-line state/finding JSON across Beads
updates. promote-legacy-feature.py attaches the canonical formula lifecycle and a validated implementation plan to an
existing roadmap root without creating a replacement epic.
Lifecycle startup uses references/SKILL-VERSION.md to record the executing installed
skill version and compare it with trustworthy local canonical evidence before mutation. Beads writes and delivery
closure follow references/INTERACTION-BOUNDARY.md: native linked-worktree
authority is shared, mutation intervals use a repository-scoped lease, foreign interaction evidence remains fail-closed,
and native Dolt publication uses scripts/guarded-beads-push.py instead of raw or force pushes.
Before executing a dstack workflow that links to a reference in this directory, read that reference completely. The calling skill remains responsible for its workflow-specific authority and completion rules.
Feature resolution
<core-dir>/scripts/resolve-feature.py resolves feature epics through Beads by canonical slug, exact or unique human
name, or ID. Use --next to select the next ready feature epic. Workflow commands should expose the canonical <slug>
reference and retain the Beads ID only for mutations and audit evidence.
Standalone executable issues use /implement-task <task-selector>. It processes exactly one open task, bug,
chore, spike, or standalone feature; feature epics and their descendants remain owned by the feature lifecycle.
verify-delivery-state.py provides the post-merge semantic check used by /close-feature to reconcile delivery claims
with the actual base-branch commit.