minhuw

This skill should be used when writing official conference-style reviews for research papers as if from a top-tier systems conference reviewer. Use when the author wants a realistic, formal peer review with scores, strengths, weaknesses, and detailed feedback following standard conference review formats.

This skill should be used when evaluating the logical flow, structure, readability, and overall quality of research paper text. Use for assessing academic writing targeting top-tier computer science conferences.

This skill should be used when performing systematic scans of research paper text to identify and fix typos, grammar errors, inappropriate words, and awkward expressions. Use for thorough proofreading and error correction in academic writing for computer science conferences.

This skill should be used when conducting comprehensive reviews of research papers for top-tier computer science conferences. Use for identifying weaknesses, unclear statements, missing evidence, and structural issues in academic papers. The skill autonomously reads files, searches for patterns, analyzes structure, and provides constructive peer-reviewer feedback.

This skill should be used when polishing academic research paper text for grammar, clarity, fluency, and natural phrasing. Specifically designed for non-native English speakers writing for top-tier computer science conferences.

This skill should be used when suggesting word or phrase alternatives for placeholders in academic research papers. Use when the author needs help selecting appropriate technical terminology for top-tier computer science conferences.

This skill should be used when providing concise summaries of research paper text. Use for quickly understanding the core content, arguments, and contributions of paper sections targeting top-tier computer science conferences.

This skill should be used when validating whether a specific word or phrase is appropriate, commonly used, and correct in academic technical contexts. Use for checking terminology in research papers targeting top-tier computer science conferences.

Execute, resume, inspect, or safely clean a validated herder-plans/ backlog as a dependency-aware multi-agent run with per-attempt token accounting. Use when the user asks to fire, run, resume, automatically complete Herder plans, clean Herder run worktrees or branches, or report a Fire run's status and token coverage. Do not use to create plans, repair plan formatting, or implement one ordinary task directly.

Survey a codebase as a senior advisor and write prioritized, self-contained Herder plans from verified repository findings without changing source code. Use when asked to audit code, find bugs or improvement opportunities, suggest evidence-backed product direction, review or reconcile existing plans, or produce a herder-plans/ backlog for $herder:fire. Route user-defined new features that require intent clarification to $herder:grill.

Initialize, validate, inspect, and manage Herder Markdown plan backlogs and execution-token ledgers. Use when creating or repairing herder-plans/, checking plan dependencies and readiness, changing plan tracking policy, inspecting execution status, reporting token coverage, or preparing plans for $herder:fire. Do not use to implement plans or orchestrate subagents.

Install or verify the implementer, reviewer, and saver agent profiles required by Herder. Use when setting up the Herder plugin in a repository, when Herder reports a missing Codex role, or when the user asks to refresh Herder's host-native agent definitions.

Validate a Herder plan directory against the canonical mechanical and semantic contracts, report plan/index/dependency/drift issues without changing files, and conservatively repair safe issues with --fix. Use when the user invokes herder:validate, asks whether herder-plans/ is Fire-ready or executable, wants a cold plan-quality audit, or asks to repair malformed, incomplete, or drifted plans. Do not use to execute plans, modify source code, or decide missing product intent.

Interview the user one decision at a time, investigate repository facts, maintain domain-model and ADR obligations, and create one validated Herder implementation plan from confirmed intent. Also refine an existing TODO or decision-blocked plan when invoked with --plan. Use when the user invokes herder:grill, wants to clarify a feature or change before implementation, asks to turn product intent into a plan, or wants to grill or stress-test an existing plan before Herder Fire executes it.

Implement CoQUIC GitHub issues end to end. Use when the user gives a GitHub issue URL/number or asks to fetch, implement, test, add Duvet/RFC annotations, update issue TODO checkboxes, comment on, or close an implementation issue in the minhuw/coquic repository.

Create consistent GitHub issues for tracking implementation work, bug fixes, design tasks, technical debt, protocol gaps, test gaps, performance work, documentation work, CI failures, interop follow-ups, and investigations in the coquic repository. Use when asked to draft, create, batch-create, organize, or update GitHub issues for things that should be fixed, implemented, reviewed later, or tracked as project work.

Fetch and fix current CodeQL and Codacy issues for the minhuw/coquic repository. Use when asked to inspect GitHub Security code scanning alerts at https://github.com/minhuw/coquic/security, Codacy current issues at https://app.codacy.com/gh/minhuw/coquic/issues/current, or to automatically repair static-analysis/security findings without weakening scanner configuration.

Debug failed remote GitHub Actions Interop workflow runs for the coquic repository. Use when given a .github/workflows/interop.yml run URL or run ID and asked to inspect CI, download failed interop logs/artifacts with gh, identify failed peer/testcase/direction combinations, reproduce them locally with interop/run-official.sh, fix reproducible coquic bugs, validate, commit, push, and dispatch a fresh Interop workflow run.

Use when answering QUIC protocol or RFC questions in this repository, especially section lookups, frames, transport parameters, ACK behavior, version negotiation, packet formats, error codes, or when grounded RFC citations are needed.

Query the Ocellus local uncore documentation RAG with Docmancer. Use when working on Intel uncore PMU events, Ocellus metrics, Grafana dashboards, CPU generation support, Linux uncore perf/sysfs behavior, Intel perfmon JSON, or any question that needs source-grounded lookup from the local `.resources/docmancer` corpus.