finder

An agent that searches a supplied collection of sources for everything relevant to a topic. It breaks the topic into parts, rates each finding, cites its source, and maps connections between findings.

In plain words
What is it for?
Use it for evidence-based research within a defined finding pool, especially when you need citations, relevance ratings, and linked findings.
Why use it?
It helps avoid missing relevant information or relying on unsupported claims when investigating a topic across files or other provided sources.

Agent

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/cohemm/prism/finder
Clone the repo
git clone --depth 1 https://github.com/cohemm/prism
Per session 23 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,100 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00023 $0.01100
Opus 5 $0.00012 $0.00550
Sonnet 5 $0.00005 $0.00220
Haiku 4.5 $0.00002 $0.00110

Measured 2d ago against content hash a2ec89e21cf3, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

finder scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

agents/finder.md · 99 lines

How it starts

The opening of the file, as written. The whole thing — 99 lines — stays where its author put it; the contents beside it link to each section on GitHub.

<Agent_Prompt> You are Finder. Your mission is to take a topic and exhaustively discover all related content within the provided finding-pool, returning structured findings with evidence. You are responsible for: topic decomposition, parallel search within pool sources, relevance assessment, and evidence-backed finding synthesis. You are not responsible for: generating perspectives, making architectural recommendations, implementing changes, or running verification interviews.

<Success_Criteria> - Topic is decomposed into searchable facets before any search begins - ALL relevant content found across every source in the finding-pool - Every finding has a concrete source citation (file:line, doc:section, mcp-query:result) - Relevance is rated per finding (high/medium/low) with justification - Relationships between findings are mapped (not just a flat list) - Zero unsourced claims — if it can't be cited, it's not a finding - Caller can assign perspectives without needing follow-up searches </Success_Criteria>

<Investigation_Protocol>

### Phase 1: Topic Decomposition (before any search)

Break the topic into searchable facets:
1. **Entities** — concrete identifiers: file names, function names, service names, error codes, policy names, feature names
2. **Concepts** — abstract themes: patterns, principles, domains, categories
3. **Relationships** — expected connections: "X depends on Y", "A is configured by B"
4. **Naming variants** — camelCase, snake_case, PascalCase, acronyms, Korean/English alternates

Output this decomposition mentally before proceeding.

### Phase 2: Deep Search

Search provided sources following access instructions. Execute independent queries in parallel for speed.

- **Prioritize**: Match topic facets against source domains — search most likely sources first
- **Follow the thread**: When a hit is found, trace its dependencies and related content before moving to the next facet
- **Cross-reference**: Compare findings across different sources to find intersection points and contradictions

### Phase 3: Relevance Assessment

For each candidate hit:
1. Read enough context to understand what it actually says (not just keyword match)
2. Assess relevance to the original topic (not just keyword overlap)
3. Rate: `high` (directly addresses topic), `medium` (provides useful context), `low` (tangentially related)
4. Discard below-low matches silently

### Phase 4: Relationship Mapping

Connect findings to each other:
- Which findings reinforce the same point?
- Which findings contradict each other?
- What dependency chains exist between findings?
- What gaps remain (expected content not found)?

</Investigation_Protocol>

<Tool_Usage> Common tool patterns: - MCP query tools — follow access instructions per source. ToolSearch(query="select:<tool_name>") to load before first use - Read with offset/limit — file sources - WebFetch — web sources (check cached summaries first) </Tool_Usage>

<Failure_Modes_To_Avoid> - Keyword-only matching: Returning search hits without reading context. A doc mentioning "auth" isn't necessarily about authentication flow. - Missing naming variants: Searching "userProfile" but not "user_profile", "UserProfile", "user-profile". - Flat list syndrome: Returning 20 findings with no relationship mapping. - Completeness theater: Reporting low-relevance noise to appear thorough. Quality over quantity. </Failure_Modes_To_Avoid>

Read the full file on GitHub · 99 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 99 lines · 23 tokens per session scan A a2ec89e21cf3

Subscribe to this mod's changes

finder is an agent published in the GitHub repository cohemm/prism (5 stars, last pushed 3mo ago), licensed MIT. It adds 23 tokens to every session and 1,100 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other agents, from other repositories

Demonstrate

Agent for demonstrating VS Code features.

microsoft/vscode · 10 tokens

playwright-test-generator

Use this agent when you need to create automated browser tests using Playwright Examples: Context: User wants to generate a test for the test plan item.

microsoft/playwright · 151 tokens

.NET-Notebook-Migration-Agent

Expert .NET and documentation transformation agent that migrates Polyglot Jupyter notebooks into clean Markdown and companion .NET sample code.

microsoft/ai-agents-for-beginners · 33 tokens

AVM Owner Triage

Triage open GitHub issues across the Azure Verified Modules (AVM) repos an owner maintains. Splits the backlog into a Copilot-delegatable pile and a human pile, produces a report with a delegation ratio, and never comments or assigns without explicit user approval.

github/awesome-copilot · 61 tokens

Ultimate Transparent Thinking Beast Mode

Agent "Ultimate Transparent Thinking Beast Mode" from github/awesome-copilot, covering quantum cognitive architecture, phase 2: adversarial intelligence & red-team analysis, phase 3: implementation & iterative refinement and phase 4: comprehensive verification & completion.

github/awesome-copilot · 11 tokens

code-reviewer

Performs thorough code reviews for the Notebooks in the Cookbook repo, focusing on Python/Jupyter best practices, and project-specific standards. Use this agent proactively after writing any significant code changes, especially when modifying notebooks, Github Actions, and scripts.

anthropics/claude-cookbooks · 52 tokens