Audit and migrate the complete Reflex example/template corpus

[0xnavarro]

Why

The public repository has the five canonical official recipes (agent-wakeup, agent-router, context-pruner, tool-risk, quality-gate) plus the minimal and Custom Reflex examples. That is not yet the complete use-case corpus discussed while defining Brida Reflex.

Before launch follow-up, reconcile the whole prior corpus deliberately instead of silently claiming every brainstormed example is already public.

Corpus to classify

General / agent operations

• alert/event gate;
• prompt-injection / security screen;
• semantic find/classification;
• public PR guard;
• lead-fit: investigate / ignore / review;
• invoice human-review gate;
• support-ticket department routing;
• gate before an expensive frontier-model/agent run;
• visual quality/control classification;
• defer with note / batch later / wake now / human escalation;
• deadline-triggered urgent escalation.

Market / research examples

• market event;
• news impact;
• asset router;
• research escalation;
• trade-risk advisory;
• portfolio alert;
• prediction-market Reflex.

Acceptance

For every item:

1. decide whether it is an official recipe, public example/template, skill reference, or intentionally deferred;
2. use synthetic/redacted fixtures only;
3. keep protected actions recommendation-only;
4. add schema/fixture validation and public-surface checks where applicable;
5. document why overlapping cases are consolidated rather than duplicated;
6. do not add hosted provider routing, credentials, tenancy internals or private Brida policy.

Track the reconciliation explicitly so the public repo can make a truthful claim about which examples/templates are available.

[0xnavarro]

Audit status after the current public corpus expansion:

• current validator reports 5 official recipe versions / 5 fixture sets / 42 Custom Reflex draft examples;
• the public corpus now covers agent progress/completion, browser/action risk, code/change review, content policy/quality, context/memory/retrieval relevance, support triage, document/invoice routing, duplicate/record matching, incident/input risk, lead/job fit, semantic cache/data/rule/response gates, research claim/event relevance, work priority, and additional bounded examples;
• workflow patterns, abstention/fan-out templates, semantic feature composition, and browser-extension semantic blocking examples are public.

This issue should remain open rather than claim full reconciliation: the explicitly listed market/news/asset/trade-risk/portfolio/prediction-market family is not yet represented as a deliberate classified set, and #3 still tracks the dedicated portable public-PR-guard/GitHub Action artifact.

Keep using this issue as the truthful corpus reconciliation ledger; do not duplicate overlapping examples merely to satisfy names.