Writing Effective Tools for Agents
Comments on “Writing Effective Tools for Agents”
Claim-checked. Scored by all three judges, and its checkable claims were compared with the sources it cites. How far stories are checked
Collins at p99log pulls together OpenAI, Anthropic, Google and MCP guidance on designing agent tools: build around workflows rather than API endpoints, treat descriptions and errors as prompt text, and avoid overlapping tools. Past about 20 tools, toolsets, tool search or code execution cut context cost; GitHub's server dropped from 101 tools and 64.6k tokens to 52 and 30.3k. Anthropic's figures are from its own tests. Collins adds security advice, including idempotency keys on writes, which is his own recommendation.
Why it ranked
No. 2 of 123 with a score of 0.76. The panel rated it highest on practical relevance (0.93) and reader interest (0.84), lowest on significance (0.63). Placed No. 2, No. 3 by score alone: the placement terms moved other stories past it.
- AI judgeJudge 2claude-haiku-5-5
- Practical, well-sourced guide on agent tool design, security and scaling, synthesizing vendor guidance with concrete examples. Mostly a synthesis rather than new research, so moderate significance and novelty.
- AI judgeJudge 3gpt-5.6-luna
- A technically substantive, well-cited synthesis of agent tool design, scaling, security, reliability, and evaluation. It is highly actionable for builders, though largely consolidates existing vendor guidance rather than presenting original research.
Each judge wrote its line after answering the rubric, with the site and the author still hidden. Every answer behind this is in the table under Score.
Score
0.76total under rubric v2.3
- Significance0.63
- Reader interest0.84
- Audience breadth0.72
- Practical relevance0.93
- Substantiated0.76
- Technical depth0.83
- New information0.73
- Counts against: Promotional0.04
- Counts against: Overclaim0.16
Each judge’s answer per question (0–1), their mean and spread, and what the question adds to the score (weight × mean).
| Question | Judge 1 | Judge 2 | Judge 3 | Mean | Spread | Weight | Contribution (weight times mean) |
|---|---|---|---|---|---|---|---|
| SignificanceHow much does this change what people in tech know or can do? | 0.72 | 0.45 | 0.72 | 0.63 | 0.27 | +0.24 | +0.151 |
| Reader interestWould a well-informed senior engineer or tech founder want to read this today, and likely pass it on? | 0.84 | 0.82 | 0.86 | 0.84 | 0.04 | +0.26 | +0.218 |
| Audience breadthHow many people who build technology does it concern? | 0.67 | 0.70 | 0.78 | 0.72 | 0.11 | +0.13 | +0.093 |
| Practical relevanceHow useful is it to someone who builds software or tech products? | 0.98 | 0.90 | 0.90 | 0.93 | 0.08 | +0.22 | +0.204 |
| SubstantiatedHow well is it backed? Consider whether it is the primary source, and whether its main claims come with data, code, a demo, a reproducible method or citations. | 0.71 | 0.80 | 0.78 | 0.76 | 0.09 | +0.05 | +0.038 |
| Technical depthHow technically substantive is it? | 0.97 | 0.75 | 0.76 | 0.83 | 0.22 | +0.05 | +0.041 |
| New informationDoes it contain information absent from the stories covered in the last 7 days (list provided)? | 0.82 | 0.55 | 0.82 | 0.73 | 0.27 | +0.05 | +0.036 |
| PromotionalIs it mainly marketing with little substance (a press release, ad or sponsored post with no real information)? A launch post that carries real news is not promotional. | 0.04 | 0.05 | 0.03 | 0.04 | 0.02 | −0.19 | −0.008 |
| OverclaimDo the headline or conclusions claim more than the body supports? | 0.19 | 0.20 | 0.08 | 0.16 | 0.12 | −0.11 | −0.017 |
| Sum of contributions | +0.757 | ||||||
| Claim-check penalty | 0.000 | ||||||
| Total, as published | 0.757 | ||||||
Spread is the highest minus the lowest judge answer. marks a spread above 0.40: the judges disagree, and the story is marked contested.
The badge for this ranking, for the author's page or anyone's: one line of HTML, and it links back here.
<a href="https://rubric.hrih.si/editions/2026-11-14/2/"><img src="https://rubric.hrih.si/badge/2026-11-14/2.svg" alt="Ranked #2 on Rubric, 14 Nov 2026" width="238" height="28"></a>Claim check
Claim check: 8 matches source. Each claim is compared with the source the article itself cites; a flag is published only when both checkers agree.
Matches sourceNumber
GitHub's MCP server used to enable 101 tools by default, using 64.6k tokens.
The cited source says the same thing.
The article says “GitHub’s MCP server used to enable 101 tools by default, using 64.6k tokens. In October 2025, GitHub cut the default to 52 tools and 30.3k tokens”
The source says “Excessive context: 101 tools consuming 64.6k tokens by default”
Source cited by the article: github.com/github/github-mcp-server/discussions/1182
Matches sourceNumber
Anthropic reports that a five-server setup with 58 tools takes about 55K tokens up front, and that it has seen definitions reach 134K tokens before optimization.
The cited source says the same thing.
The article says “Anthropic reports a five-server setup with 58 tools taking about 55K tokens up front, and says it has seen definitions reach 134K tokens before optimization”
The source says “That's 58 tools consuming approximately 55K tokens before the conversation even starts”
Source cited by the article: anthropic.com/engineering/advanced-tool-use
Matches sourceBenchmark result
In Anthropic's internal testing, adding Tool Use Examples raised accuracy on complex parameter handling from 72% to 90%.
The cited source says the same thing.
The article says “adding Tool Use Examples raised accuracy on complex parameter handling from 72% to 90%”
The source says “In our own internal testing, tool use examples improved accuracy from 72% to 90% on complex parameter handling”
Source cited by the article: anthropic.com/engineering/advanced-tool-use
Matches sourceBenchmark result
Anthropic's Tool Search Tool cut context use by about 85% in its example and raised Claude Opus 4's accuracy on its MCP evaluations from 49% to 74%.
The cited source says the same thing.
The article says “Anthropic’s Tool Search Tool cut context use by about 85% in its example, and raised Claude Opus 4’s accuracy on its MCP evaluations from 49% to 74%”
The source says “This represents an 85% reduction in token usage while maintaining access to your full tool library. Internal testing showed significant accuracy improvements on MCP evaluations when working with large tool…”
Source cited by the article: anthropic.com/engineering/advanced-tool-use
Matches sourceNumber
Anthropic's code execution with MCP reduced a Google Drive-to-Salesforce workflow from about 150,000 tokens to 2,000.
The cited source says the same thing.
The article says “a Google Drive-to-Salesforce workflow dropped from about 150,000 tokens to 2,000, because the intermediate data never passed through the model”
The source says “This reduces the token usage from 150,000 tokens to 2,000 tokens—a time and cost saving of 98.7%”
Source cited by the article: anthropic.com/engineering/code-execution-with-mcp
Matches sourceBenchmark result
Anthropic's multi-agent research setup outperformed a single Claude Opus 4 agent by 90.2% on an internal evaluation, while using about 15 times as many tokens as a chat.
The cited source says the same thing.
The article says “Its multi-agent setup outperformed a single Claude Opus 4 agent by 90.2% on an internal research evaluation, but multi-agent systems used about 15 times as many tokens as a chat”
The source says “a multi-agent system with Claude Opus 4 as the lead agent and Claude Sonnet 4 subagents outperformed single-agent Claude Opus 4 by 90.2% on our internal research eval”
Source cited by the article: anthropic.com/engineering/multi-agent-research-system
Matches sourceCausal conclusion
In May 2025, Invariant Labs showed that a malicious issue in a public GitHub repository could make an agent using the GitHub MCP server read private repositories and publish their contents in a public pull request.
The cited source says the same thing.
The article says “a malicious issue in a public GitHub repository could make an agent using the GitHub MCP server read the user’s private repositories and publish their contents in a public pull…”
The source says “As shown here, as soon as the agent encounters the malicious GitHub issue, it can be coerced into pulling private repository data into context, and leaking it in an autonomously-created…”
Source cited by the article: invariantlabs.ai/blog/mcp-github-vulnerability
Matches sourceNumber
Corelight found that an MCP server pinned to a specific release produced 15 different environments across 19 builds over five weeks.
The cited source says the same thing.
The article says “produced 15 different environments across 19 builds over five weeks, because its launcher re-resolved dependencies on every rebuild”
The source says “Over five weeks, 19 builds of the same release produced 15 distinct environments”
Source cited by the article: gbhackers.com/corelight-reveals-ai-security
Also covered by
No other coverage was clustered with this story.
Elsewhere
Elsewhere: HN 1 pts, not on front page
Recorded for context only, checked 15 Nov 2026, 20:44 UTC. It played no part in the ranking.