construct-daily-cycle

Runs a non-blocking daily cycle composing research, curation, and graph-health capabilities into one digest.

Updated Apr 21, 2026
One-click install
npx skills add https://github.com/mabstruct/construct --skill construct-daily-cycle-mabstruct
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: construct-daily-cycle
Source: https://github.com/mabstruct/construct/tree/main/CONSTRUCT-CLAUDE-impl/claude/skills/construct-daily-cycle
Command: npx skills add https://github.com/mabstruct/construct --skill construct-daily-cycle-mabstruct

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Keeping a personal knowledge workspace current requires running research, curation, and graph-health checks every day, which is tedious to coordinate manually. This Skill runs the entire daily cycle as a single non-blocking command and narrates the composed digest, so you get caught up without babysitting each step. ## Core Features & Use Cases - Composed Daily Run: Invokes the construct daily run CLI (or the construct_daily_run MCP tool), which chains the frozen Python research.run, curation.run, and graph.status capabilities in one pass. - Honest Digest Narration: Reports the run-level status (completed, degraded, or failed), each child cycle's message, and the closing graph-health summary using the capability's own counts. - Escalation Surfacing: Reports the pending-escalation count of items the gates deliberately did not auto-decide, and points you to construct research review and construct curation review for interactive handling. - Use Case: You open your workspace in the morning and say "catch me up" — the Skill runs the full daily cycle and returns a digest covering new research, curation outcomes, and knowledge-graph health, plus a count of items awaiting your review. ## Quick Start Ask the agent to run the daily cycle by saying "run the daily cycle" or "catch me up" in a workspace where the construct CLI is on your PATH.

Frequently Asked Questions about construct-daily-cycle

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run the daily cycle in a CONSTRUCT workspace?▼

Say "daily", "run the daily cycle", or "catch me up" to the agent, which invokes `construct daily run --workspace . --json`. The command composes research, curation, and graph-status in a single non-blocking pass and returns one DailyRunResult.

What does the daily digest from construct daily run include?▼

The digest includes a run-level status (completed, degraded, or failed), per-child results for research.run, curation.run, and graph.status, a closing graph-health summary with card, connection, and domain counts, and a pending-escalation count.

Can I approve or reject items during the daily run?▼

No. The daily run is non-blocking and auto-applies each child gate's recommended decision, never auto-writing escalated items. To review pending items, run `construct research review` or `construct curation review` interactively afterward.

Does the daily cycle skill write directly to the workspace?▼

No. The skill holds no direct web-fetch or workspace-write capability; every write flows through the Python daily.run capability behind its own gates. Views regeneration also happens in the Python layer, gated on views/build/ and views.auto_regenerate.

What does a degraded status mean in the daily run result?▼

A degraded status means a child cycle failed or degraded, or there are pending escalations left undecided. It is reported plainly in the digest and is never collapsed into a bare completed status.