agentsclimarketplace

Harvest feed

Skill mnttnm/claude-skills/harvest-feed

Custom Claude Code skills spanning UI/UX, productivity, and developer tooling

Install
npx -y skills add mnttnm/claude-skills --skill harvest-feed

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Mine the current conversation for publishable feed entries for the digital garden site. Use when the user says 'harvest', 'harvest feed', 'mine this session', 'extract posts', 'what can I publish from this session', or at the end of a work session to turn learnings into content. Analyzes chat history to find non-obvious tricks, useful discoveries, tooling insights, and project progress — then proposes 4-5 feed entries ready to publish.

SKILL.md

6.3 KB, as published. Nobody here has run it

Harvest Feed

Scan the current conversation history and generate 4-5 feed entry suggestions for the digital garden at ~/work/projects/digital-garden/.

What Makes a Good Entry

Prioritize these (highest to lowest):

  1. Novel techniques — something done in a non-obvious way that others would find useful
  2. Tooling discoveries — how Claude Code, MCP servers, CLI tools, or AI workflows were used effectively
  3. Gotchas and fixes — problems hit and how they were solved, especially if the solution wasn't obvious
  4. Project progress — meaningful milestones, shipped features, or architectural decisions (changelog-style)
  5. Workflow insights — patterns for working with AI that improved speed or quality

Skip anything generic that a reader could easily find in official docs.

Procedure

1. Scan and Extract

Review the full conversation. For each candidate, note:

  • What happened (the concrete thing)
  • Why it's interesting (the non-obvious part)
  • Who benefits from knowing this (the audience signal)
  • What visual artefact could make this entry more engaging (see step 2b)

2a. Classify Each Entry

Read references/content-formats.md for exact schemas.

Pick the best type for each:

SignalType
Quick observation, one insight, 2-5 sentencesDiscovery (learning)
A useful link/resource worth sharingDiscovery (resource)
Progress on an active projectProject activity entry

For project activity entries, check which project file to append to:

ls ~/work/projects/digital-garden/src/content/projects/

2b. Suggest Visual Artefacts

For each entry, suggest one or more artefacts that would make the post more engaging. Think about what the reader would want to see alongside the text.

Types of artefacts to suggest:

ArtefactWhen to suggestExample
ScreenshotUI changes, tool output, terminal results, before/after comparisons"Screenshot of the Shopify admin after product import"
Code snippetA specific command, config block, or script that made things workUse the code + codeLanguage fields
Before/after pairMigrations, redesigns, refactors"Screenshot of WooCommerce vs Shopify product page"
Terminal outputCLI tool results, build logs, error messages that tell the storyScreenshot or code block of the relevant output
Diagram or flowArchitecture decisions, data pipelines, workflow explanations"Simple flow: WooCommerce export → script → Shopify CSV → upload"
Short screen recordingMulti-step workflows, interactive features"Record the /harvest-feed flow from invocation to file creation"

Rules:

  • Every entry should have at least one artefact suggestion, even if it's just "Screenshot of [specific thing]"
  • Be specific about what to capture — "Screenshot of the terminal" is too vague, "Screenshot of the Shopify CLI theme dev output showing hot reload" is good
  • If the artefact already exists in the conversation (e.g. user shared a screenshot earlier), reference it
  • Suggest where to save images: public/captures/ (or public/images/projects/[project-slug]/ for project activity)

3. Present Suggestions

Present all 4-5 suggestions in a numbered list. For each show:

## [N]. [Title]
**Type**: Discovery (learning) / Discovery (resource) / Project update → [project-slug]
**Why it's worth sharing**: One sentence on the novelty or usefulness.
**Suggested artefact**: What to capture and where to save it.

[Draft content — full markdown ready to paste into a file]

4. Ask for Selection

After presenting all suggestions, ask:

Which of these should I create? (e.g. "1, 3, 5" or "all" or "skip 2") For entries with suggested screenshots — do you want to capture those now, or add them later?

5. Write Files

For approved entries, create the actual files:

  • Discovery: src/content/discoveries/YYYY-MM-DD-slug.md
  • Project activity: Append to the activity: array in the existing project file

Use today's date. Set draft: true so nothing publishes until the user explicitly flips it.

If artefacts are code snippets, include them directly using code and codeLanguage fields.

For image artefacts, add to the images array with the expected path. Leave a TODO note if the screenshot still needs to be captured.

Writing Style

REQUIRED: Read references/writing-style.md before drafting any content. It defines voice, titles, punctuation rules, structure, and what to avoid.

Critical rules (always apply, even without loading the guide):

  • Always wrap frontmatter title values in double quotes
  • Limit em dashes to 1-2 per post. Vary with periods, commas, parentheses
  • Lead with what it is, not why you're writing about it
  • No lecture tone, no takeaway wrap-ups, no salesy framing

Reference Examples

Before writing, review these examples to match the tone and structure of existing posts:

ExampleTypeWhat it demonstrates
examples/learning-with-images.mdLearningMulti-tool comparison, inline images, personal take as blockquote
examples/resource-concise.mdResourceAudience-targeted intro, concise, no claims
examples/learning-long-form.mdLearningTechnical depth, sections, prompt inclusion
examples/resource-minimal.mdResourceOne-liner commentary, minimal
examples/resource-premise-driven.mdResourcePremise-first framing, context before link

Match the closest example to each entry you draft.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.