agentleFS
priyanshu-aryapriyanshu-arya/LinkedIn_Automation/.claude/skills/generate-ideas/SKILL.md

generate-ideas

Skill0 starsChanged 2 months agoMITView on GitHub

Use when the user asks to generate LinkedIn content ideas from the research backlog, run the Idea Ranker/Idea Engine, or explicitly invokes /generate-ideas. Turns unused Research Notes into scored, ranked candidate Idea Notes in Post-Ideas/. Does not assign ideas to specific days or draft post text — that's the Content Strategist / LinkedIn Writer (later phase).

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 4 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Default target range is 10–20, but this is a ceiling, not a quota: only generate from research notes that are actually usable.

    line 21
  2. Very specific:

    Read frontmatter across Content-Research//*.md (optionally filtered to domain-filter).

    line 29
  3. Very specific:

    Before turning a research note into an idea, scan Post-Ideas/*.md (frontmatter topic/angle, any status except placeholder) for close …

    line 35
  4. Distinctive:

    Never fabricate an idea from a research note that doesn't exist or is a placeholder.

    line 92
  5. Very specific:

    Never invent facts beyond what the source research note actually supports — an idea's why_it_matters and hook must be grounded in the …

    line 94
  6. Distinctive:

    Never assign a specific posting day/date — that's out of scope here.

    line 97
  7. Distinctive:

    Never write LinkedIn post copy — ideas only, not drafts.

    line 98

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Generate Ideas (Idea Ranker — Phase 3)

Converts unused research into a pool of scored, ranked candidate LinkedIn post ideas, written to Post-Ideas/ using _Templates/Idea-Note.md. This does not pick what gets posted this week or write any post text — it only builds and ranks the candidate pool. Content Strategist (a later phase) consumes this pool to plan the week; LinkedIn Writer (a later phase) drafts the actual posts.

Arguments

Parse from the invocation (e.g. /generate-ideas AI 5):

  • domain-filter (optional) — restrict to one category folder under Content-Research/. If omitted, consider all categories.
  • count (optional) — target number of new ideas to generate this run. Default target range is 10–20, but this is a ceiling, not a quota: only generate from research notes that are actually usable. If the backlog can only support fewer good ideas, generate fewer and say so in the report — never pad with filler ideas to hit a number.

Process

1. Gather eligible research

Read frontmatter across Content-Research/**/*.md (optionally filtered to domain-filter). Eligible = status: new. Explicitly exclude status: placeholder and status: used — placeholders are hand-written schema examples, never real input.

2. Dedup against existing ideas

Before turning a research note into an idea, scan Post-Ideas/*.md (frontmatter topic/angle, any status except placeholder) for close matches. If a near-duplicate idea already exists, either give the new idea a materially different angle or skip that research note this run.

3. Generate one idea per eligible research note

Fields (per REQUIREMENTS.md §6): topic, angle, why_it_matters, target_audience, format, hook, estimated_engagement (0-10, your judgment of how engaging this specific angle is), suggested_visual, category, content_type, sources (the research note id(s) actually used), target_week (leave blank — Content Strategist fills this in).

You may synthesize 2 closely-related research notes into a single comparison-style idea when it is clearly stronger than two separate ideas (e.g. an "X vs Y" format) — but only then, and only citing sources actually used. Do not force synthesis to hit a count target.

4. Score rank_score (0-10)

Average these five, each scored 0-10:

FactorWhat it measures
Hook StrengthIs the hook specific and attention-grabbing, or generic?
Educational/Practical ValueWill the reader walk away with something concrete?
Audience FitRelevance to AI/dev/career LinkedIn audience
Non-repetition / FreshnessHow different is this from existing ideas/published posts (from the dedup check in step 2)
Format-Content FitDoes the chosen format actually suit this material?

This is a static formula. Once Phase 10 (Growth Agent) has real published-post performance data, expect these weights to be replaced with learned ones — don't treat this rubric as permanent.

5. Write the Idea Note

Copy _Templates/Idea-Note.md into Post-Ideas/ as YYYY-MM-DD--kebab-slug.md. Fill in all fields from steps 3-4. status: candidate.

6. Close the loop on used research

For every research note used as a source in step 3:

  • Append the new idea's id to that research note's used_in[].
  • Flip its status from new to used. This is what makes Phase 2's dedup check (skip topics already turned into ideas) actually work over time.

Report back

  • List each new idea: filename, topic, category, content_type, rank_score.
  • A diversity breakdown of the full candidate pool in Post-Ideas/ (not just this run's additions): counts by category and by content_type, across all status: candidate ideas (excluding placeholder). If the pool is thin or skewed toward one category, say so plainly — that's a signal more /research-topic runs are needed before Content Strategist can build a properly diverse week, not something to paper over.

Hard rules

  • Never fabricate an idea from a research note that doesn't exist or is a placeholder.
  • Never invent facts beyond what the source research note actually supports — an idea's why_it_matters and hook must be grounded in the research note's Summary/Key Findings, not embellished.
  • Never assign a specific posting day/date — that's out of scope here.
  • Never write LinkedIn post copy — ideas only, not drafts.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.