new-research

Featured

Start a durable research / source note that captures what you read and found — fires on "start a research note", "capture these sources", "log my research". Part of the Agentsmith harness; scaffolds docs/research/<slug>.md so findings live in the repo, never in disposable memory (R9), and are never silently deleted.

AI & Automation 175 stars 6 forks Updated 2 days ago MIT

Install

View on GitHub

Quality Score: 89/100

Stars 20%
75
Recency 20%
100
Frontmatter 20%
70
Documentation 15%
69
Issue Health 10%
50
License 10%
100
Description 5%
100

Skill Content

# New research note Research belongs in the repo, traceable to sources, and is never silently deleted (R9). This scaffolds one note per topic. ## When this fires "start a research note" / "capture these sources" / beginning any multi-source investigation whose findings should outlive the session. ## Fast path — if the Agentsmith CLI is available Run `agentsmith new-research "topic name"` — use `.agentsmith/agentsmith` on macOS/Linux or `.agentsmith\\agentsmith.cmd` on Windows when the command is not on PATH. It slugifies the title and writes `docs/research/<slug>.md` from the template (it won't overwrite an existing file). ## Fallback — no CLI 1. slug = the topic lowercased, every run of non-alphanumeric characters → `-`, ends trimmed. 2. If `docs/research/<slug>.md` already exists, STOP — append to it, never overwrite. 3. Otherwise create it with a no-delete banner (obsolete → move to `docs/research/_archive/`, R9) and these sections: - **Question / scope** - **Sources consulted** — url or citation; what you actually opened, dated. - **Findings** — each traceable to a source above; mark single-sourced / uncertain items. - **Open questions / what was NOT checked.** ## Report Name the file created and remind the operator to commit it (R9 — research lives in git, not memory). Every finding you add later must cite a source already in the note.

Details

Author
PromptPartner
Repository
PromptPartner/agentsmith
Created
2 months ago
Last Updated
2 days ago
Language
Python
License
MIT

Similar Skills

Semantically similar based on skill content — not just same category