| name | companies-like-x |
| description | Given one seed company, find peers along an explicit similarity axis (product, customer, tech stack, business model, culture). Writes to <WORKING_FOLDER>/domain-notes/like-<seed-slug>.md. Reads outreach.md to mark contact status. Use when the user has a known reference point and wants more of the same shape. |
| disable-model-invocation | false |
| allowed-tools | Read, Write, WebSearch, WebFetch, Bash(mkdir *), Bash(test *) |
Companies Like X
Given a seed company, enumerate peers along a chosen similarity axis. Avoids generic "competitors of X" — forces the user to pick the dimension.
Inputs
$ARGUMENTS:
- Required:
<seed> — company name or domain.
- Optional:
--axis=<axis> — one of product, customer, stack, business-model, culture. Default: ask.
- Optional:
--n=<count> (default 15).
- Optional:
--exclude-direct-competitors — skip head-to-head rivals; surface adjacent peers instead.
- Optional:
--refresh.
Examples
/career:discover companies-like-x snowglobe.ai --axis=product
/career:discover companies-like-x linear --axis=culture --n=20
/career:discover companies-like-x retool --axis=customer --exclude-direct-competitors
Procedure
1. Resolve config + paths
${CAREER_DATA_DIR}/config.json → WORKING_FOLDER. Output: ${WORKING_FOLDER}/domain-notes/like-${seed-slug}.md.
2. Profile the seed
Single-pass research on <seed>:
- One-line product description.
- Customer segment (ICP).
- Tech stack signals (job postings, engineering blog).
- Business model (b2b saas, marketplace, prosumer, open-core, services).
- Culture signals (remote posture, values, comp band, public hiring tone).
If --axis not provided, surface the five axes with a one-liner each and ask the user to pick before continuing.
3. Enumerate peers along the axis
WebSearch + WebFetch around the chosen axis only. Examples:
product: companies solving the same job for the user.customer: different products serving the same buyer.stack: companies using the same core tech (e.g., "Rust + Postgres + b2b API").business-model: open-core peers, prosumer-saas peers, etc.culture: remote-first, async-default, transparent comp, etc.
Aim for --n matches. Drop matches that only fit on a different axis (note them in "Out of axis").
4. Cross-reference outreach.md
Same as discover-by-domain step 4. Mark each peer with contact-status.
5. Write the note
Format:
# Companies like {{seed}} — axis: {{axis}}
> Seed: {{seed}}
> Axis: {{axis}}
> Surveyed: {{date}}
> Peers enumerated: {{n}}
## Seed profile
- Product: ...
- Customer: ...
- Stack: ...
- Business model: ...
- Culture: ...
## Why this axis
{{1–2 sentences on what "similar on {{axis}}" means here}}
## Peers
| name | website | why-similar (one line) | stage | hq | remote? | contact-status | notes |
| --- | --- | --- | --- | --- | --- | --- | --- |
## Out of axis (similar on a different dimension)
- {{name}} — similar on {{other-axis}}, not {{chosen-axis}}.
## Sources
1. ...
6. Print summary
peers note: domain-notes/like-<seed-slug>.md
axis: <axis>
peers: <n>
never-contacted: <n>
contacted-stale: <n>
contacted-recent: <n>
next: /career:suggest companies --from=like-<seed-slug>
Guardrails
- One axis per run. Re-run with a different
--axis for a different lens — don't blur axes.
- "Why-similar" must be specific to the axis, not generic ("also a SaaS company" is not a reason).
- No fabrication. If you can't find evidence for the claimed similarity, drop the row.
Failure modes
- Seed unknown / no signal → bail; ask the user for a website or a clarifying note.
- Axis ambiguous ("similar but better") → ask the user to pick from the five axes.
