Browse all
Hexagonal Architecture with InversifyJS DI Pattern
This skill guides the implementation of Hexagonal Architecture using InversifyJS in a monorepo, defining clear boundaries between core domain logic, ports, and external adapters. It provides recipes for wiring dependencies, managing scopes,…
Universal Content Extraction and Summarization Tool
This utility extracts text content from a wide array of sources, including URLs, PDFs, media files, and web pages. It supports advanced features like structured data extraction and LLM-powered summarisation.
ThumbGate: Safety guardrails for AI agents
ThumbGate provides pre-action gates for AI coding agents by intercepting tool calls to prevent known mistakes. It captures structured feedback and automatically generates prevention rules from repeated failure patterns.
Secure IPC for Vault Credential Management
Provides a unified Unix socket for secure, authenticated communication with a local vault. It facilitates sensitive operations like payment signing and credential retrieval while ensuring private keys are never stored long-term.
Workflow for adding new MCP tools
This guide details the mandatory workflow for integrating a new Market Connector Protocol (MCP) tool. It covers implementation, registration, pricing, and unit testing within the established codebase structure.
Capture structured agent feedback and rules
This skill processes explicit user quality signals (e.g., thumbs up/down) by capturing detailed context, failure modes, and suggested prevention actions. It structures this feedback into reusable memories and rules for improving future agen…
Bluesky Reply Monitoring and Draft Queueing
This skill monitors the AT Protocol notifications endpoint for replies, mentions, and quotes to queue human-reviewable drafts. It avoids auto-posting by writing drafts to a JSONL file for manual approval or downstream consumption.
Automated diagnosis of complex system architecture issues
This skill diagnoses and manages complex operational failures, covering areas from architectural flaws and deployment errors to autonomy and ROI concerns. It guides the agent to utilize specialized tools for systematic debugging and process…
Persistent Agent Memory and Failure Prevention
This tool provides persistent, structured memory for agents across sessions. It allows developers to capture detailed success or failure feedback, automatically generating prevention rules that guide the agent to avoid repeating past mistak…
Recall Historical Coding Context
Retrieve historical failure patterns, prevention rules, and bundled context packs to avoid repeating past mistakes during coding tasks.
Migrates Agent Markdown Memory to Enforcement System
This utility bridges the agent-architect-kit's markdown memory structure to the ThumbGate feedback system. It parses per-role markdown files, converting structured learnings and mistakes into actionable, queryable feedback signals for runti…
Search lessons for corrective actions and rules
This tool queries the promoted lessons database to retrieve comprehensive information regarding failure patterns. It searches for corrective actions, linked prevention rules, gates, and the current lifecycle state of the lessons.
Agent Run Governance and Guardrails
This skill provides comprehensive governance for programmatic agent runs, ensuring controlled execution across local, cloud, or self-hosted environments. It enforces pre-flight checks, scope bounding, and evidence capture before agents can …
Structured Task Feedback Capture
Records structured thumbs up/down feedback with context, tags, and rubric scores, featuring automated lesson distillation from chat history to drive prevention rule promotion.
Auto-generate and manage failure prevention rules
This tool manages prevention rules, which are automatically generated when specific failure patterns recur. Developers can list active rules, check if an action is blocked, or review the history of rule promotion.
Validate the Codex bridge plugin setup
This skill validates the readiness of the Codex bridge plugin within Claude Code. It checks for the installation status of `codex`, the availability of `codex exec review`, and verifies the presence of necessary artifact directories and ser…
ThumbGate Feedback Capture Skill
Captures explicit user feedback into structured memories and prevention rules to facilitate agentic learning. It logs context, failures, and corrective actions to refine future agent performance.
Retrieve Codex Bridge Artifact Metadata
Retrieves the latest metadata from the Codex bridge, including execution mode, timestamp, exit status, and artifact paths. This allows for auditing recent bridge executions and locating generated artifacts.
Codex Code Review Bridge
Executes Codex reviews against uncommitted changes, specific branches, or commits using the Claude Code bridge. Supports targeted analysis by passing focus areas via prompts.
Second pass review using Codex bridge
This skill facilitates an independent second review of the current task or repository state by handing the context to Codex via a dedicated bridge script. It is designed for explicit user requests, allowing the agent to synthesize and prese…
Retrieve last saved Codex bridge result
This skill allows developers to inspect the raw, previously saved output from the Codex bridge process. It prevents unnecessary re-execution of the underlying model while providing immediate access to the latest artifact.
ThumbGate Dual-Write Feedback Skill
Captures user feedback simultaneously in Amp MCP memory for in-session recall and thumbgate for DPO export, analytics, and cross-platform prevention rules.
Run skeptical adversarial code reviews
Utilises a Codex bridge to perform a deep, skeptical review of code changes, identifying potential regressions, unsafe assumptions, and release risks. This is ideal for pre-merge checks, migrations, or any high-risk deployment.
Self-learning memory layer for coding agents
Instinct is a local memory layer that observes coding patterns, tracking confidence through repeated usage. It automatically promotes recurring fixes and preferences into actionable rules and suggestions for future sessions.