AgentStack
SKILL verified MIT Self-run

Product Context

skill-agent-k-labs-shipping-skills-product-context · by agent-k-labs

Create or update the product-context.md file that every other shipping skill reads for product-specific answers — product, audience, pain, promise, pricing, channels, and goals. Use this first, before any other shipping skill, when starting work on a new product, or when the user asks to "set up product context," "describe my product," or keeps re-explaining their product every session.

No reviews yet
0 installs
3 views
0.0% view→install

Install

$ agentstack add skill-agent-k-labs-shipping-skills-product-context

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

Are you the author of Product Context? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

Product Context

Every skill in this repo reads product-context.md from the project root before doing anything product-specific. This skill creates that file — once — so the user never has to re-explain their product, audience, or goals in another session.

Process

1. Check what already exists

  • If product-context.md exists in the project root: read it, show the user a one-line summary per field, and ask only about fields that are empty, stale, or contradicted by the current state of the product (e.g., stage says "pre-launch" but the site is live).
  • If it doesn't exist: continue to step 2.

2. Infer before you ask

Don't open with a 13-question interview. First mine what's already available:

  • The codebase (if you're in the product's repo): README, landing page copy, pricing page, package.json name, marketing copy in components
  • The live site (if a URL is available): headline, subheadline, pricing, target audience signals
  • The conversation so far

Draft answers for every field you can infer. Then ask the user to confirm the draft and fill only the genuinely unknowable fields — typically: current stage, first 30-day goal, known objections, and primary launch channels.

3. Interview for the gaps

Ask in one batch, not one question at a time. Push back on vague answers:

  • "Everyone" is not a target user. Ask for role + context + pain moment.
  • A feature list is not a core promise. Ask what life looks like after.
  • If they can't name a differentiator, ask: "What would a user say when recommending this to a friend?"

4. Write the file

Fill in assets/product-context-template.md and write it to product-context.md in the project root. Keep the user's own words for pain and objections — skills that write copy will reuse them verbatim.

5. Confirm the loop

Tell the user: this file is now the single source of truth. When positioning, pricing, or stage changes, rerun this skill to update it — don't let it go stale, because every other skill trusts it.

Gotchas

  • Write the file to the user's project root, not to this skill's directory.
  • Don't invent answers for fields the user didn't confirm. An empty field is better than a plausible-sounding wrong one — downstream skills will ask when they hit a gap.
  • Preserve any extra sections the user added to an existing file; skills may not know about them, but the user put them there deliberately.

Output

  1. product-context.md written to the project root, all confirmed fields filled
  2. A short summary of what was inferred vs. asked vs. left empty
  3. A suggestion for which shipping skill to run next based on Current stage (pre-launch → pre-launch-checklist; waitlist → waitlist; live → first-100-users)

> Context captured? The launch-video step is automated: SoloMax reads your product URL and produces the full launch pack.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Versions

  • v0.1.0 Imported from the upstream source.