Edition of Wednesday 18 November 2026

Microsoft's agent-framework on PyPI shipped a credential stealer in two releases

Comments on “Microsoft's agent-framework on PyPI shipped a credential stealer in two releases”

Claim-checked. Scored by all three judges, and its checkable claims were compared with the sources it cites. How far stories are checked

Aikido Security reports that agent-framework 1.26.1 and 1.26.2, Microsoft's Python AI agent meta-package on PyPI, were malicious releases published 33 minutes apart on Nov 17 and removed by about 21:20 UTC. A .pth hook ran a stealer on every Python start, collecting Azure, Azure DevOps and GitHub Actions secrets and sending them to webhook.site. How the publishing credentials were obtained is unknown. Most claims rest on Aikido's own analysis. Anyone who installed either version should pin 1.26.0 and rotate reachable secrets.

Read the original at aikido.dev. The judges saw only its text, with the site, the author and link domains removed; they are shown again after ranking.

Why it ranked

No. 1 of 142 with a score of 0.80. The panel rated it highest on practical relevance (0.95) and new information (0.93), lowest on audience breadth (0.65).

AI judgeJudge 2claude-haiku-5-5
Detailed, primary-level analysis of a live supply-chain credential stealer in a widely used AI SDK, with hashes, code excerpts and IOCs; highly practical for builders, though it includes vendor promotion and a limited audience scope.
AI judgeJudge 3gpt-5.6-luna
A detailed, actionable supply-chain compromise analysis with code excerpts, affected versions, indicators, hashes, and remediation guidance. It is highly relevant to builders, though the inserted vendor promotion reduces neutrality.

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

0.80total under rubric v2.3

Judges’ mean answer per question, on a scale from 0 to 1. Questions marked minus count against the score.
  • Significance0.76
  • Reader interest0.92
  • Audience breadth0.65
  • Practical relevance0.95
  • Substantiated0.88
  • Technical depth0.93
  • New information0.93
  • Counts against: Promotional0.20
  • Counts against: Overclaim0.14

Each judge’s answer per question (0–1), their mean and spread, and what the question adds to the score (weight × mean).

QuestionJudge 1Judge 2Judge 3MeanSpreadWeightContribution (weight times mean)
SignificanceHow much does this change what people in tech know or can do?0.730.720.820.760.10+0.24+0.182
Reader interestWould a well-informed senior engineer or tech founder want to read this today, and likely pass it on?0.890.920.950.920.06+0.26+0.239
Audience breadthHow many people who build technology does it concern?0.630.700.620.650.08+0.13+0.085
Practical relevanceHow useful is it to someone who builds software or tech products?0.980.950.920.950.06+0.22+0.209
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.0.880.930.820.880.11+0.05+0.044
Technical depthHow technically substantive is it?1.000.920.880.930.12+0.05+0.047
New informationDoes it contain information absent from the stories covered in the last 7 days (list provided)?0.850.970.980.930.13+0.05+0.047
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.0.060.200.350.200.29−0.19−0.039
OverclaimDo the headline or conclusions claim more than the body supports?0.220.120.080.140.14−0.11−0.015
Sum of contributions+0.798
Claim-check penalty0.000
Total, as published0.798

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.

Ranked #1 on Rubric, 18 Nov 2026
<a href="https://rubric.hrih.si/editions/2026-11-18/1/"><img src="https://rubric.hrih.si/badge/2026-11-18/1.svg" alt="Ranked #1 on Rubric, 18 Nov 2026" width="238" height="28"></a>

Claim check

Claim check: 1 matches source · 7 no source given. Each claim is compared with the source the article itself cites; a flag is published only when both checkers agree.

  1. No source givenDate

    Microsoft Agent Framework's agent-framework PyPI meta-package versions 1.26.1 and 1.26.2 were malicious releases published 33 minutes apart on Nov 17, 2026, running a credential stealer targeting Azure, Azure DevOps, and GitHub Actions secrets.

    The article gives no link or citation for it.

    The article says

    “two malicious releases of Microsoft Agent Framework's PyPI meta-package published 33 minutes apart on Nov 17, 2026”

    The source says

    No source to compare with.

  2. No source givenNumber

    The agent-framework package sees about 190,000 downloads a month.

    The article gives no link or citation for it.

    The article says

    “The package sees about 190,000 downloads a month”

    The source says

    No source to compare with.

  3. No source givenDate

    Version 1.26.1 appeared at 20:35 UTC with a credential stealer that runs every time Python starts, whether or not the package is imported, and sends secrets to a webhook.site URL.

    The article gives no link or citation for it.

    The article says

    “At 20:35 UTC, version 1.26.1 appeared with a credential stealer that runs every time Python starts, whether or not anything imports the package”

    The source says

    No source to compare with.

  4. No source givenDate

    By 21:20 UTC both malicious versions had been removed from PyPI and 1.26.0 was the latest release again.

    The article gives no link or citation for it.

    The article says

    “By 21:20 UTC both versions had been removed and 1.26.0 was the latest release again”

    The source says

    No source to compare with.

  5. Matches sourceCausal conclusion

    Python's site module executes any line beginning with import in every .pth file in site-packages at startup.

    The cited source says the same thing.

    The article says

    “any line that begins with import gets executed”

    The source says

    “For backward compatibility, lines starting with import (followed by space or tab) are executed with exec”

    Source cited by the article: docs.python.org/3/library/site.html

  6. No source givenVersion

    Version 1.26.2 added theft of GitHub Actions OIDC tokens, which 1.26.1 did not have, and raised the credential file cap from 60 to 80.

    The article gives no link or citation for it.

    The article says

    “GitHub Actions OIDC theft: no in 1.26.1, yes in 1.26.2”

    The source says

    No source to compare with.

  7. No source givenDate

    The LiteLLM compromise in March 2026 used the same .pth startup hook to run its payload on every Python launch.

    The article gives no link or citation for it.

    The article says

    “The LiteLLM compromise in March 2026 used the same startup hook to run its payload on every Python launch”

    The source says

    No source to compare with.

  8. No source givenCausal conclusion

    The attacker likely unpacked a legitimate wheel, edited it by hand, and uploaded the repacked file with valid publishing credentials.

    The article gives no link or citation for it.

    The article says

    “That fits someone who unpacked a legitimate wheel, edited it by hand, and uploaded the repacked file with valid publishing credentials”

    The source says

    No source to compare with.

Also covered by

No other coverage was clustered with this story.

Elsewhere

Elsewhere: HN 4 pts, not on front page

Discussion on Hacker News

Recorded for context only, checked 18 Nov 2026, 05:47 UTC. It played no part in the ranking.

Panel for this edition

Four AI agents ranked this edition: three judges scored every story blind to its site and author, and a fourth compared the top stories’ claims with the sources they cite. Summaries are written by AI.

Model versions are pinned for every edition. How ranking works describes rubric v2.3.

Comments

Readers’ comments on this story have a page of their own, ranked by a public rubric rather than by votes.

Comments on “Microsoft's agent-framework on PyPI shipped a credential stealer in two releases”

More on Rubric

From the last 7 editions, chosen by shared topic, score and date.