Skip to main content
Glama
forest-USTC

FeatureLens Research

Official
by forest-USTC

Server Configuration

Describes the environment variables required to run the server.

NameRequiredDescriptionDefault
FEATURELENS_DATA_DIRNoThe data directory used by FeatureLens. Defaults to `~/.featurelens/data`; can also be specified with the `--data-dir` command-line option.~/.featurelens/data

Instructions

Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.

This server publishes no instructions, or was last inspected before Glama recorded them.

Capabilities

Features and capabilities supported by this server

Protocol revision2025-11-25

CapabilityDetails
tools
{
  "listChanged": false
}
prompts
{
  "listChanged": false
}

Tools

Functions exposed to the LLM to take actions

NameDescription
research_createB

建立产品功能研究,保存用户的具体问题和明确选择的目标。target 为可选的已解包 JavaScript 目录 绝对路径;不自动寻找其他应用。

research_listA

列出研究,默认仅活动项;status 支持 active、archived、deleted、all。

research_getA

查看研究问题、判断、验证记录、分析范围和证据目录;完整证据使用 research_evidence 读取。

research_analyzeA

对当前研究明确指定的 已解包 JavaScript 目录运行内置语法扫描器,保存完整原始 Evidence。最多 180 秒、64 MiB 输出,进程 512 MiB 堆;仅标准 JS,不支持 JSX/TS/ASAR/.app。只分析代码,不执行目标。大应用优先提供 source_paths 缩小范围。不会自动产生业务结论。

research_import_analysisB

导入用户选择的完整 兼容的 JavaScript 分析 Evidence JSON,最多 64 MiB;导入文件不授权读取其内部引用的本地路径。

research_find_nodesA

按字面关键词搜索完整静态分析图并分页;query 不是语义搜索。指定 node_id 可读取完整节点及覆盖限制。节点不等于实际运行行为。

research_capture_nodeA

将已选择的静态节点登记成可引用证据,保留原始 Evidence ID、节点 ID 与哈希。同一节点重复登记会复用证据。

research_read_sourceB

读取静态线索对应的少量源码行,先核对文件哈希。仅支持本工具实际分析过的、用户指定的已解包目录,拒绝路径越界和文件变更。

research_add_evidenceA

保存研究者提供的资料、静态代码、运行观察或访谈记录。必须写明来源;工具不会把提交的文字当成已独立核验的事实。

research_evidenceB

按 ID 查看证据,或按字面关键词检索研究内已登记的证据并分页。

research_add_claimB

记录研究者的材料事实 fact、分析推断 inference 或待确认事项 unknown。fact/inference 必须引用本研究证据;inference 必须说明局限。引用检查不证明结论正确。

research_add_verificationA

为待确认问题创建可执行的验证项,包含步骤、预期结果和关联判断;此工具只登记计划,不执行目标。

research_record_verificationA

记录人工验证的实际观察与通过/失败/阻塞状态。通过和失败必须引用证据;保留历次尝试,不自动升级任何判断。

research_reportA

汇编带引用和局限的 Markdown 或 JSON 报告。默认返回内容;指定 output 绝对路径时写入新文件,不覆盖现有文件。报告是研究记录,不是模型自动核验。

research_doctorA

检查本工具版本、数据目录和内置分析器信息,不修改配置,不检查或安装原生反编译引擎。

research_import_documentB

提取本地 TXT/MD/DOCX/文字型 PDF,保留文件哈希与来源位置;无 OCR,不执行文件指令,不调用模型。

research_document_chunksB

分页或字面搜索文档片段,返回页码、段落或行号;检查原件和提取结果哈希,不是 RAG。

research_capture_document_chunkB

将选定文档片段登记成可引用证据,保留来源位置和文件哈希;不验证原文真实性。

research_updateA

修改研究标题或问题并保留修订历史,不改写已有证据和判断。

research_archiveA

归档研究,从默认列表隐藏;保留材料,可取消归档。

research_unarchiveB

取消归档,让研究重新出现在默认列表。

research_deleteA

将明确选择的研究移入回收区;保留原始证据,可通过 undelete 恢复。

research_undeleteB

从回收区恢复研究及原归档状态。

research_backupA

将整个研究库与引用的原始材料打包为 ZIP,包含归档和回收区;不覆盖已有文件。

research_restoreA

校验 ZIP 备份并恢复到尚不存在的新目录;不覆盖当前库,不授权读取历史应用路径。

Prompts

Interactive templates invoked by user choice

NameDescription
feature-investigation产品功能研究:问题、证据、判断、验证与报告。

Resources

Contextual data attached and managed by the client

NameDescription

No resources

TDQS

A3.5/5.0

Scored across 25 tools

Disambiguation4/5

Tools target mostly distinct resources/actions, and descriptions carefully differentiate e.g. add_verification (plan) vs record_verification (outcome) and capture_document_chunk vs capture_node vs add_evidence. However, the three evidence-capture tools and evidence/get overlap conceptually, which could still cause misselection without careful reading.

Naming Consistency4/5

All tools share the research_ prefix and use snake_case throughout, giving a predictable family. The convention is mostly verb_noun, but a few are noun-only (research_evidence, research_doctor, research_document_chunks), a minor deviation.

Tool Count3/5

25 tools sits at the heavy boundary flagged by the rubric; the lifecycle (archive/unarchive/delete/undelete/backup/restore) adds several low-traffic tools. The count is defensible for a full research-workflow domain but leans bloated.

Completeness4/5

The surface covers a full research lifecycle: create/get/update/list, archive/delete with recovery, backup/restore, analysis, document import, evidence, claims, verification, and reporting. Gaps remain in per-item editing/removal (no update/delete for evidence or claims), which agents may need to work around.

Maintenance

ActivityMaintained
ResponsivenessNo issues