extract-artifact-candidates

Mines chat history for durable task-local candidates and backfills ARTIFACT_CANDIDATES.md files.

2|Updated Apr 29, 2026
One-click install
npx skills add https://github.com/siarhei-belavus/agent-public --skill extract-artifact-candidates-siarhei-belavus
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: extract-artifact-candidates
Source: https://github.com/siarhei-belavus/agent-public/tree/main/skills/atelier-workflow/extract-artifact-candidates
Command: npx skills add https://github.com/siarhei-belavus/agent-public --skill extract-artifact-candidates-siarhei-belavus

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Long implementation or debugging sessions often produce valuable decisions, rejected approaches, and hidden constraints that never get recorded before the conversation fades. This Skill performs a manual salvage pass over the current chat history to capture that durable residue into a task packet's ARTIFACT_CANDIDATES.md file. ## Core Features & Use Cases - History Mining: Scans the current conversation and context logs for non-obvious rationale, hidden constraints, owner-boundary notes, rejected paths, revisit triggers, and start-here breadcrumbs. - Candidate Triage: Applies a strict candidate bar so only durable, costly-to-forget facts are kept, filtering out debugging noise, todo residue, and implementation summaries. - Packet Backfill: Appends or merges compact candidates into .plans/<task>/ARTIFACT_CANDIDATES.md and notes the harvest in INDEX.md, without touching tracked AGENTS.md or PLAN.md. - Use Case: After a long debugging session where several approaches were rejected in chat but never written down, run this Skill to backfill those rejected paths and constraints into the task packet before they are lost. ## Quick Start Ask the assistant to mine the current conversation for forgotten durable candidates and backfill them into the current task packet's ARTIFACT_CANDIDATES.md.

Frequently Asked Questions about extract-artifact-candidates

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

FAQPage Schema
How do I recover forgotten decisions from a long AI chat session?▼

Run a manual history harvest that scans the current conversation for non-obvious rationale, rejected approaches, and hidden constraints, then writes them as compact candidates into the task packet's ARTIFACT_CANDIDATES.md file.

What kinds of chat content should be saved as artifact candidates?▼

Keep non-obvious rationale, hidden constraints, owner-boundary clarifications, rejected tempting approaches, revisit triggers, and start-here navigation hints. Reject implementation summaries, transient debugging noise, todo residue, and anything already obvious from code or tests.

What happens if no task packet exists when harvesting candidates?▼

The Skill does not invent a packet silently. It reports the candidates in chat and recommends bootstrapping a new packet via the plan-task workflow instead.

Does this Skill update AGENTS.md or PLAN.md files?▼

No. It only writes to ARTIFACT_CANDIDATES.md and adds a short harvest note in INDEX.md. Tracked AGENTS.md, AGENTS.override.md, PLAN.md, and AMENDMENTS.md are explicitly out of scope.

When should I not run a candidate harvest?▼

Avoid it when the packet status is already done and the candidate inbox is complete, or when the chat only contains routine implementation detail. It is a salvage pass, not a replacement for the distill-knowledge workflow.