suggest-projects

name	suggest-projects
description	Suggest concrete projects (build / write / contribute) the user could ship to close a skill gap, demonstrate a capability, or create a hook for outreach to a target company. Writes to <WORKING_FOLDER>/recommendations/projects-<date>.md.
disable-model-invocation	false
allowed-tools	Read, Write, WebSearch, WebFetch, Bash(mkdir *), Bash(test *), Bash(date *)

Suggest Projects

Translates "what should I work on" into concrete shipped artifacts. Each project has a clear scope, time estimate, and what it would unlock (which gap it closes, which conversation it starts).

Inputs

$ARGUMENTS:

• Optional: --close-gap=<skill> — close a specific skill gap.
• Optional: --target-company=<slug> — produce a project that's a credible hook for outreach to this company.
• Optional: --target-domain=<slug> — domain-shaped project (signal-building).
• Optional: --time-budget=<weekend|week|month> (default week).
• Optional: --n=<count> (default 5).

Examples

/career:suggest projects --close-gap="RLHF tooling" --time-budget=weekend
/career:suggest projects --target-company=snowglobe --time-budget=week
/career:suggest projects --target-domain=ai-evals --n=8

Procedure

1. Load context

• ground-truth.md.
• Source signals: skills note, company briefs, domain note.

2. Generate candidates

For each candidate project, define:

• Shape: build / write / contribute / present.
• Concrete artifact: repo / blog post / OSS PR / talk / dataset / template.
• Time estimate: maps to --time-budget.
• What it demonstrates: ties back to a specific skill or domain claim.
• What it unlocks: gap closed, hook for outreach, portfolio piece.
• Distribution: where the artifact lives + first 3 places to share it.

3. Filter

• Drop anything that needs more than 2× the time budget.
• Drop anything that requires private/proprietary access the user doesn't have.
• Keep diversity: don't suggest 5 blog posts; mix shapes.

4. Write the recommendations file

# Project suggestions — {{date}}

> Goal: {{close-gap | target-company | target-domain}}
> Time budget: {{budget}}
> Generated: {{datetime}}

## Projects

### 1. {{title}}
- shape: {{build|write|contribute|present}}
- artifact: {{repo|post|PR|talk|dataset}}
- time: {{est}}
- demonstrates: {{skill / claim}}
- unlocks: {{gap closed | outreach hook | portfolio}}
- distribution: {{where to publish + share}}
- starter steps:
  1. ...
  2. ...

### 2. ...

## Why these and not others

- {{1–2 sentences on the trade-offs you made}}

## Sources

1. ...

5. Print summary

recommendations: recommendations/projects-<date>.md
projects: <n>

next: pick one and commit to a time slot in /career:plan

Guardrails

• A project must be shippable in the time budget. "Write a paper on X" without scoping is not a project.
• Hook-shaped projects must have a real connection to the target. Don't manufacture a fake project to "have something to send".
• Don't suggest closed-source replicas of the target's product as a hook — it reads as competitive, not collaborative.

Failure modes

• Time budget < gap closing effort → recommend a smaller scope or extend budget.
• No target signal → ask for --close-gap or --target-domain.