ihabkhaled
UserAI-Psychiatry is a Claude plugin that gives AI coding agents an executive-function layer to detect and control attention drift, overthinking, recursive reasoning, compulsive verification, scope drift, retry loops, deadlocks, livelocks, context waste, and failure to stop after completion.
Categories
Indexed Skills (50)
anti-gaming
Use when commands, labels, agents, counters, or classifications change while the same underlying behavior continues.
blocker-validator
Use when BLOCKED is proposed after difficulty, uncertainty, unfamiliar code, a slow operation, or a small number of failed attempts.
completion-evidence
Use when DONE is proposed with unrun tests, missing requirements, inferred success, happy-path-only proof, or unresolved required findings.
context-balance
Use when context is overloaded or compression may have removed requirements, constraints, evidence, blockers, or decisions.
decision-readiness
Use when an important implementation, architecture, security, data, or delivery decision rests on unresolved critical unknowns.
evidence-floor
Use when mandatory behavior, integration, security, migration, or data requirements lack the appropriate kind of proof.
executive-override
Use when an AI-Psychiatry budget expires while new evidence shows a narrow extension is required for correctness, security, or completion.
false-progress-detector
Use when status reports emphasize tools, files, commits, plans, tokens, or agents while requirements and evidence remain unchanged.
framework-red-team
Use when an agent-control framework needs adversarial testing for fake productivity, loops, recursion, evidence gaps, or vague wording.
hidden-recursion-detector
Use when nested work is renamed, moved across agents, promoted to top level, or hidden behind research, validation, review, and dependency labels.
investigation-floor
Use when non-trivial implementation starts before what exists, what changes, why, risks, and validation are sufficiently known.
loophole-hunter
Use when rules appear satisfied while their intent is bypassed, counters reset, classifications change conveniently, or an AI-Psychiatry audit is requested.
memory-validator
Use when durable memory may contain assumptions, stale facts, contradictions, duplicates, obsolete decisions, or temporary debugging state.
reasoning-balance
Use when a task may be underthinking, sufficiently reasoned, or overthinking and the correct next action is unclear.
root-cause-validator
Use when debugging changes assertions, mocks, exceptions, or symptoms without explaining materially important failure behavior.
rule-conflict-resolver
Use when system, user, repository, domain, AI-Psychiatry, skill, memory, or task-state instructions appear incompatible.
scope-laundering-detector
Use when adjacent work is called required without dependency proof, or inconvenient required work is marked optional.
strategy-laundering-detector
Use when different commands, tools, runners, plans, or agents repeat the same hypothesis and evidence target.
underthinking-detector
Use when implementation, parking, blocking, strategy switching, or completion happens before critical understanding and evidence exist.
all-the-medicine
Use when the user explicitly requests the complete AI-Psychiatry framework through one all-in-one command that evaluates and orchestrates every available skill until the task is proven complete.
direct-communication
Use when answers, progress updates, explanations, or questions risk verbosity, repetition, indirect wording, vague conclusions, or conversational loops.
never-stop
Use when the user explicitly requests maximum autonomous execution, no unnecessary questions, persistent recovery, independent decision-making, and continuous work until the full Definition of Done is proven.
anti-overthinking
Use when analysis paralysis, speculative branching, repeated replanning, premature architecture work, or indecision blocks a reversible implementation or test.
attention-reset
Use when the agent changes objectives, chases optional discoveries, context-switches repeatedly, abandons the active item, or shows ADHD-like distraction as a behavioral analogy.
bounded-investigation
Use when debugging, architecture inspection, research, or repository exploration risks becoming open-ended, recursive, speculative, or disconnected from a decision.
communicate-briefly
Use when reporting work, progress, failures, retries, tests, blockers, files, or completion and the message risks verbosity, filler, repetition, or hidden critical information.
completion-gate
Use when requested behavior appears complete, remaining work is optional, perfectionism delays delivery, or the agent continues auditing, refining, testing, or researching after proof.
context-compression
Use when a long session, large tool output, repeated discussion, or handoff needs a compact state that preserves decisions and enables continuation.
context-refresh
Use when active context is large, stale, contradictory, repeatedly reread, or no longer represents current source and task state.
context-router
Use when choosing which repository instructions, architecture, domain knowledge, files, tests, or deep guidance to load for a task without reading everything.
critic-controller
Use when reviewers or judges repeat rounds, invent requirements, block on style, expand scope, or continue after correctness, security, regression, and explicit requirements are satisfied.
deadlock-recovery
Use when no productive next action can be selected because dependencies, authority, contradictory constraints, missing access, or unresolved evidence block progress.
evidence-gate
Use when files, services, commands, tests, APIs, configuration, architecture, publication, or runtime results are assumed without an authoritative repository or tool source.
executive-function
Use when a task needs explicit control of objective, scope, priority, nesting, retries, verification, progress, drift, recovery, or termination.
failure-learning
Use after a non-trivial, recurring, or expensive failure when a concise evidenced lesson could prevent future repeated investigation.
goal-lock
Use when the requested outcome is unclear, silently changing, being reinterpreted by discoveries, or disconnected from the active implementation.
handoff-task
Use when work moves to another agent, session, person, or environment and the recipient needs enough state to continue without a sprawling narrative.
knowledge-maintainer
Use when repository architecture, rules, skills, commands, context routing, manifests, or generated AI knowledge change and dependent sources may become stale.
livelock-recovery
Use when activity is high but requirements, blockers, tests, and deliverables remain unchanged across repeated edit, search, test, review, or planning cycles.
loop-detector
Use when commands, searches, file reads, edits, tests, critics, plans, context reloads, or strategies repeat without new evidence or acceptance progress.
memory-curator
Use when deciding whether task discoveries, decisions, preferences, architecture, failures, or lessons deserve durable repository memory.
multi-agent-coordinator
Use when two or more agents work concurrently, delegation may recurse, responsibilities overlap, dependencies can deadlock, or a coordinator must preserve scope and termination.
progress-audit
Use when status is unclear, activity is high, completion estimates feel unreliable, or the agent may be mistaking reading, reasoning, tools, or churn for progress.
repository-map-update
Use when packages, services, ownership boundaries, entrypoints, tests, commands, or dependencies changed and the compact repository map needs a focused update.
resume-task
Use when continuing work from saved task state, a compacted session, a handoff, or a new agent without rereading the entire repository.
scope-guard
Use when newly discovered work, technical debt, refactors, audits, optimizations, or reviewer suggestions threaten to expand the user-authorized task.
strategy-reset
Use when three same-strategy attempts fail, two approaches oscillate without evidence, or the current abstraction and evidence source cannot resolve the blocker.
targeted-testing
Use when selecting development and final validation commands so evidence is sufficient without repeatedly running expensive or irrelevant suites.
task-bootstrap
Use when beginning any plan, audit, investigation, modification, review, test, or expensive tool action before the primary objective and completion conditions are explicit.
verification-controller
Use when deciding what evidence is required, whether previous proof remains valid, or whether repeated testing has become reassurance rather than necessary verification.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.