← 返回 Skills 市场

Pronoun Resolver

Name: Pronoun Resolver
Author: kaicianflone

作者 Kai Cianflone · GitHub ↗ · v0.9.0 · MIT-0

cross-platform ⚠ suspicious

总下载

当前安装

版本数

在 OpenClaw 中安装

/install pronoun-resolver

功能描述

Intercepts ambiguous pronouns (it, them, these, those, that, this, they, its) in user prompts and resolves them to specific referents using a tiered LLM engi...

使用说明 (SKILL.md)

Pronoun Resolver

Automatically detects ambiguous pronouns in your prompts and resolves them to specific referents before Claude acts on them.

How It Works

Every message is scanned for pronouns: it, them, these, those, that, this, they, its
If the prompt is self-contained (pronoun + specific noun), it passes through untouched
If a pronoun is ambiguous, a quick LLM self-check resolves it with a confidence score
If confidence is low, 3 independent LLM agents vote on the resolution
The resolved referent is injected as context — your original message is never modified

Ledger

Resolutions are logged to .claude/pronoun-ledger.json in the project directory. The ledger tracks accuracy and adjusts the confidence threshold over time:

High accuracy: trusts self-check more (fewer council escalations)
Low accuracy: escalates to council more often

Disable

Create .claude/pronoun-resolver-disabled in the project root to disable. Delete the file to re-enable.

Install

Symlink or copy this directory to ~/.claude/skills/pronoun-resolver:

ln -s /path/to/coding-pronoun-prompt-resolver ~/.claude/skills/pronoun-resolver

安全使用建议

Install only if you are comfortable with every prompt being scanned, ambiguous prompts and correction checks being sent through local Claude CLI model calls, and prompt-derived resolution history being stored in a project ledger. Add the ledger to .gitignore, review or purge it periodically, and use the disable file in sensitive projects.

能力标签

financial-authoritycan-make-purchasesrequires-sensitive-credentials

能力评估

⚠ Purpose & Capability

The pronoun-resolution purpose is coherent, but the skill operates on broad ambiguous commands such as fix, update, delete, and rename by injecting silent resolution context before the assistant acts. Publisher metadata also lists unrelated high-risk capability tags, which do not match the artifact behavior.

⚠ Instruction Scope

The hook runs on every user prompt when installed, performs correction monitoring on follow-up prompts, and can invoke additional LLM stages without per-prompt confirmation. A project-level disable file exists, but default activation is broad.

ℹ Install Mechanism

Installation is a standard skill copy or symlink with a user-prompt-submit command hook. Static scan and dependency scan are clean, and VirusTotal telemetry is absent, but the hook grants automatic execution on all prompts.

⚠ Credentials

For pronoun resolution, reading each prompt and invoking an LLM is expected, but passing project paths and asking the model to use project context expands exposure beyond the immediate prompt.

⚠ Persistence & Privilege

The skill stores a plain JSON ledger under .claude with prompt excerpts, inferred referents, timestamps, confidence values, and correction state. It documents pruning and gitignore guidance, but persistence is default and not opt-in.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install pronoun-resolver
安装完成后，直接呼叫该 Skill 的名称或使用 /pronoun-resolver 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v0.9.0

Eval framework (10 cases), resolve.py extraction, adversarial review fixes (stdin guard, fence stripping, sed injection, backtick matching)

v0.1.0

Initial release: tiered LLM resolution engine with self-learning ledger

元数据

Slug pronoun-resolver

版本 0.9.0

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 2

常见问题