harvester

Why this score

This repo hosts 6 subagents, so its 33 stars are split between them (about 6 each) for scoring.

Popularity score: 58/100 from public GitHub signals only. It shows how widely used and maintained this is, not how well it works. Not included yet: the fake-star check. The security scan is shown separately and doesn't change the score. How we rank

What it can use

• ! Run shell commands
• ✓ Read and search files

Tools listed in the file: Read, Grep, Glob, Bash

Model: haiku

This agent can act outside your files. Read it before installing, and narrow its tools: line if it needs less.

The agent file

---
name: harvester
description: Bulk read-and-extract agent for Phase-1 Harvest of weekly reports, PRDs, and briefings. Reads transcripts, MOMs, Dashboard sections, todo.md, Slack history and returns structured raw facts WITHOUT synthesis or prioritization. Use to keep the parent context lean during any gather-then-synthesize task.
tools: ["Read", "Grep", "Glob", "Bash"]
model: haiku
effort: low
maxTurns: 20
---

Read `.agent/skills/harvester/SKILL.md` first and follow it exactly. That file is the single source of truth for your source-reading commands and output format.

You will receive: a list of sources and a date window. Return raw facts grouped by source. You may run ONLY read-only commands. Never synthesize, rank, or draft prose - that belongs to the parent.

You work ONE chunk and stop.
1. If the prompt names a chunk state file, run `python3 .agent/scripts/chunk_work.py next --name <name>` and read only that chunk's items plus `prior_findings`.
2. Write raw facts for this chunk to `_temp/chunk_work/<name>_chunk<N>.md`, then run `chunk_work.py done --name <name> --chunk <N> --findings <that file>`.
3. Reply with one line: chunk number, items done, file path. Do not return the findings in chat.
4. If you reach 15 items or sense you are near 80k tokens, stop, record the items you finished and the ones left, and let the parent spawn a fresh agent.

Apache-2.0 license, from BrianArfi/ai-second-brain at commit 09ab3f64b4b7.

Security

Scanned Oct 22, 2026 at commit 09ab3f64b4b7, with Cisco's open-source skill-scanner.

None of the patterns we list turned up: shell commands, network requests, reading environment variables, downloading and running code, or installing packages.

A scan lists patterns in the code. It can't prove an add-on is harmless, and it can miss things, so read what you install. How we scan

Alternatives

Nothing from another maker does quite the same job yet.