skills/selective-intelligence/references/no-paid-capability-recovery.md

# No-Paid Capability Recovery

Selective Intelligence adapts to the AI, account, files, tools, permissions, and access the person already has. The person does not adapt their life or budget to the system.

When a required capability is absent, restricted, or paywalled, use this order:

1. inspect the current environment and use an existing capability the person already has;
2. inspect `metadata/no-paid-capabilities.json` and reuse or compose the bundled open tools;
3. use a reputable free or open standard/tool that fits the existing environment and preserves the same authority, privacy, and truth rules;
4. when the task authorizes implementation, build the smallest complete reusable open utility, test it against the real outcome, register it in the catalog, and make it available to later users;
5. if the truthful outcome still cannot be delivered, report the exact missing capability and retain it as an improvement frontier instead of prescribing payment.

Capability recovery is outcome-based. Match the job the person needs done, not the branding or visual surface of a commercial product. Preserve evidence for supported features, limitations, security boundaries, operating cost, and maintenance burden. A no-cost implementation that shifts hidden hosting, key, configuration, or maintenance work onto a non-developer has not passed.

AI companies control their own tier quotas, models, sign-in rules, tools, and service boundaries. Treat those as external constraints that Selective Intelligence must respect, not defects it can override. Use only legitimate available routes. If those boundaries prevent the wanted result, retain the wanted result as open and report the exact block; do not silently substitute a smaller outcome or call the person satisfied.

Never:

- bypass a paywall, authentication, quota, license, or technical protection;
- copy proprietary code, protected content, trade dress, or a closed service;
- use an unofficial or unsafe route merely because it is free;
- claim full parity when only a bounded job or subset has been reproduced;
- require the person to change AI clients, buy a plan, supply a provider key, run commands, or manage infrastructure as the default resolution.

The catalog is a machine-readable starting point, not a marketing claim. A capability is listed only when its implementation path ships in the canonical release. Live or client-specific equivalence still needs executed evidence.