growth-engine
Skill2 starsChanged 3 days agoMITView on GitHub
Turns everything the search produces into growth: a knowledge graph of what the candidate can prove, a market map of what employers ask for, a ranked skill-gap report, proof-project and learning recommendations, positioning and targeting advice from the funnel, and a 30-minute weekly review with a plan. Use for 'skill gaps', 'what should I learn', 'what is the market asking for', 'why am I not getting interviews', 'weekly review', or when the funnel stalls.
6 other projects use a version of this.
Adopted by
Most recently taken up by
- LazyIsEfficient/agentic-ostheir own version, first seen Oct 22, 2026
- agentik-os/claude-code-skillstheir own version, first seen Sep 17, 2026
- ericosiu/ai-marketing-skillstheir own version, first seen Sep 2, 2026
- KameronKales/planfi-skillstheir own version, first seen Jul 3, 2026
- varunk130/ai-marketing-claude-skillstheir own version, first seen May 18, 2026
- admdanielpaulino/claude-codestheir own version, first seen Apr 9, 2026
Other versions6 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 6
- ericosiu/ai-marketing-skills★ 3.8ksame name
- LazyIsEfficient/agentic-os★ 17same name
- admdanielpaulino/claude-codes★ 1same name
- agentik-os/claude-code-skills★ 1same name
- varunk130/ai-marketing-claude-skills★ 1same name
- KameronKales/planfi-skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository13 other files it gives its agents.
- skills/agent-setup/references/agents.mdAGENTS.md
- agent-setupSkill
- application-kitSkill
- apply-assistSkill
- candidate-intakeSkill
- inbox-syncSkill
- interview-coachSkill
- job-discoverySkill
- job-trackerSkill
- networkingSkill
- offer-coachSkill
- self-improveSkill
- until-hiredSkill
The file
We split this skill into the rules it gives an agent
5 rules. 0 are common, 2 are distinctive, and 3 are very specific to this project.
0
Common
Most of the 6 edited versions we read keep these.
2
Distinctive
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
- Very specific: line 37
When a proof is done: file {"skill": "<skill>", "proof": "<link>"}, then knowledge.py learn --status done --json <path>, then a quick …
- Very specific: line 52
Write it to knowledge/reviews/<YYYY-MM-DD>.md, update PREFERENCES.md / config.json with the decisions, and run knowledge.py summary --write.
- Very specific: line 58
Never move a skill to evidenced without a proof the candidate can point to.
- Distinctive: line 59
Never present recommendations as guarantees, or salary research as financial advice.
- Distinctive: line 60
Never turn the review into pressure: the candidate sets the pace.
“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Growth engine
Applying is half the job. The other half is getting more hireable every week, guided by real evidence instead of guesses.
The knowledge graph (<workspace>/knowledge/)
Built and kept by python3 <workspace>/bin/knowledge.py:
- Nodes: skills (with a level: evidenced, claimed, learning, none), projects, roles the candidate scored, gaps, learning resources.
- Edges: project uses skill · role requires skill (must-have or nice-to-have) · gap about skill · skill learning_from resource.
- Market map (
market.json): how often each skill appears in the job descriptions scored so far, and how often it is a must-have, per track. It grows by itself: intake adds skills and projects, every scored JD adds demand, every debrief or rejection reason adds gaps, every finished proof moves a skill to evidenced.
1. Gap report
knowledge.py gaps --write → knowledge/GAPS.md, ranked by (times asked + times required + 3 × interview mentions):
| Class | Meaning | Default recommendation |
| STRENGTH | evidenced and in demand | lead with it in the resume summary, notes and interviews |
| DEEPEN | a strength interviewers still probed | deeper answer, one sharper story, a mock round |
| UNPROVEN | claimed, nothing to show | highest interview risk: a small public proof, or stop claiming it |
| LEARNING | in progress | finish the proof |
| MISSING | in demand, not theirs | close it only if it fits the target track; otherwise filter those jobs out |
| INTERVIEW | raised in interviews, not a JD keyword | honest answer + practice |
| Trust the shares only after about 20 scored JDs; say so when there are fewer. |
2. Recommendations (write them in knowledge/RECOMMENDATIONS.md, newest first)
For the top 3–5 gaps, decide with the candidate:
- Close it or route around it? Weigh demand share in their target track, distance from what they already know, and time to a proof.
- Proof project for gaps worth closing: the smallest public piece of work that an interviewer would accept as evidence, done in 1–2 weeks. Spec: goal, scope, the measurable result, what to publish (repo, write-up, demo), and the exact line it adds to
FACTS.mdwhen done. - Learning plan: 2–3 resources (official docs and courses first), hours per week the candidate can really give.
- Honest wording until then ("I've used X in a side project; in production I've done Y").
When a proof is done: file
{"skill": "<skill>", "proof": "<link>"}, thenknowledge.py learn --status done --json <path>, then a quick candidate-intake update soFACTS.mdgains the line (signed off by the candidate).
Also recommend, from the evidence:
- Positioning: which strengths to lead with; headline and summary wording (drafts the candidate edits; never edit their public profiles yourself).
- Targeting: from
tracker.py funnel, which sources, tracks and score bands produce responses and interviews; propose more of those and less of the rest. Needs about 30 applications before the numbers mean much. - Adjacent roles where their evidenced skills already match the market well.
3. Weekly review (about 30 minutes, on cadence.weekly_review_day)
Prepare it before asking for the candidate's time:
- Numbers:
tracker.py funnelanddashboard: applied, responses, interviews, offers this week and overall; median days to response. - What worked / what didn't: by source, track and score band; interview debriefs this week.
- Gaps: top 5 from the refreshed
GAPS.md, with movement since last week. - Decisions for the candidate (they decide; you write them down): learning goal and proof project, targeting changes, limits, platforms to add or pause, score weights.
- Plan for next week: applications target, interviews and prep, learning hours, one profile improvement.
- Check-in: "How are you holding up? Anything to change about pace?" Adjust the plan to their energy; a sustainable search beats a burnt-out one.
Write it to
knowledge/reviews/<YYYY-MM-DD>.md, updatePREFERENCES.md/config.jsonwith the decisions, and runknowledge.py summary --write.
4. Market research (monthly, or when targets change)
With web search, if available: hiring trends and typical pay for the candidate's target roles and locations, with sources and dates. Add findings to knowledge/MARKET.md. Treat it as context, not advice; the tracker's own funnel beats general trends.
Never
- Never move a skill to
evidencedwithout a proof the candidate can point to. - Never present recommendations as guarantees, or salary research as financial advice.
- Never turn the review into pressure: the candidate sets the pace.
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
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.

