acquisition-graduation

Acquisition Graduation

Use this skill after dev-browser solves a non-trivial source acquisition and the path appears reusable.

Do not graduate one-off browsing. Graduate only when the acquisition path is likely to recur and can be validated without secrets or brittle session state.

Graduation Criteria

All must be true:

• repeated source or domain class,
• stable URL, API, selector, or CDP/network pattern,
• success survived at least two acquisition attempts or has obvious deterministic structure,
• validation checks can detect drift,
• legal and rate-limit constraints are known,
• no access-control bypass,
• no secrets, cookies, tokens, or session artifacts are needed.

What To Store

Store the map, not the diary:

• source/domain,
• task class,
• URL patterns,
• selectors or CDP/network observations,
• hidden API endpoints if discovered legally,
• waits, redirects, traps,
• screenshots or preservation steps,
• validation checks,
• legal/rate-limit notes,
• what not to automate.

What Not To Store

Never store:

• passwords, cookies, bearer tokens, or session IDs,
• personal accounts,
• CAPTCHA bypass paths,
• fragile pixel coordinates,
• one-off narration,
• screenshots that expose secrets,
• instructions to evade rate limits or access controls.

Output Location

For now, write reusable acquisition notes under:

skills/integrations/references/acquisition-paths/{domain-or-source}.md

If the path becomes broad enough to deserve its own skill, propose a new skill rather than silently expanding this one.