docs/ai/README.md

# AI Framework Guidance

Start with [human behavior analogies](human-behavior-analogies.md) and the [failure-mode catalog](../../.ai/guides/failure-mode-catalog.md).

Focused interventions:

- [ADHD-like attention drift analogy](../../.ai/guides/attention-drift.md)
- [OCD-like compulsive checking analogy](../../.ai/guides/compulsive-verification.md)
- [Overthinking and analysis paralysis](../../.ai/guides/overthinking-analysis-paralysis.md)
- [Box-inside-box recursive investigation](../../.ai/guides/recursive-investigation.md)
- [Nested job and subagent control](../../.ai/guides/nested-job-control.md)
- [Hallucination and evidence control](../../.ai/guides/hallucination-evidence.md)
- [Perfectionism and completion avoidance](../../.ai/guides/perfectionism-completion.md)
- [Rabbit holes and scope drift](../../.ai/guides/rabbit-holes-scope-drift.md)
- [Deadlock, livelock, and strategy oscillation](../../.ai/guides/deadlock-livelock.md)
- [Semantic compliance and anti-gaming](../../.ai/guides/semantic-compliance.md)
- [Loophole catalog](../../.ai/guides/loophole-catalog.md)
- [Underthinking detection](underthinking.md)
- [Reasoning balance](reasoning-balance.md)
- [Context starvation](context-starvation.md)
- [Sufficient reasoning](sufficient-reasoning.md)
- [Executive overrides and rule conflicts](../../.ai/guides/executive-override-conflicts.md)

Explicit-invocation superpowers:

- [Relentless execution](relentless-execution.md) and [autonomous decision-making](autonomous-decision-making.md), backing [`never-stop`](../../.ai/guides/relentless-execution.md).
- [Approval boundaries](approval-boundaries.md) — what proceeds automatically and what requires authorization.
- [All the medicine](all-the-medicine.md), backing [`all-the-medicine`](../../.ai/guides/all-the-medicine.md).

Operational authority remains in the canonical master prompt, focused `.ai/rules/`, and callable skills. Load deep guidance only when its observable trigger occurs.