Edition of Thursday 19 November 2026

Microsoft's agent-framework PyPI package Compromised with Credential Stealer

Comments on “Microsoft's agent-framework PyPI package Compromised with Credential Stealer”

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

Socket Research reports that versions 1.26.1 and 1.26.2 of Microsoft's agent-framework package on PyPI were compromised with credential-stealing code. Both versions use Python .pth startup hooks to silently harvest Azure, GitHub Actions, and CI/CD credentials and send them to a webhook.site endpoint.

Read the original at socket.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. 117 of 144 with a score of 0.60. The panel rated it highest on practical relevance (0.93) and reader interest (0.92), lowest on new information (0.55). The judges disagreed most on new information (spread 0.92), so the story is marked contested. The claim check took 0.200 off. Placed No. 117, No. 90 by score alone: the 33rd Security story of the day (−0.080), and close to “Microsoft's agent-framework on PyPI shipped a credential stealer in two releases”, No. 1 on Wednesday 18 November 2026 (−0.024). The placement terms move stories; they are not part of the score.

AI judgeJudge 2claude-haiku-5-5
Detailed, well-substantiated supply-chain compromise of a widely used AI framework package with code-level analysis and IOCs; highly relevant to builders handling credentials and CI, though a single incident.
AI judgeJudge 3gpt-5.6-luna
A highly actionable, technically detailed supply-chain compromise affecting CI/CD and cloud credentials. Its substance is strong, but it duplicates a story explicitly listed as covered in the last seven days.

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

0.60total under rubric v2.3Contestedjudges differ by 0.92 on new information

Judges’ mean answer per question, on a scale from 0 to 1. Questions marked minus count against the score.
  • Significance0.69
  • Reader interest0.92
  • Audience breadth0.65
  • Practical relevance0.93
  • Substantiated0.89
  • Technical depth0.91
  • New information0.55
  • Counts against: Promotional0.02
  • Counts against: Overclaim0.09

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?0.690.600.780.690.18+0.24+0.165
Reader interestWould a well-informed senior engineer or tech founder want to read this today, and likely pass it on?0.890.920.960.920.07+0.26+0.240
Audience breadthHow many people who build technology does it concern?0.610.720.620.650.11+0.13+0.084
Practical relevanceHow useful is it to someone who builds software or tech products?0.970.900.910.930.07+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.0.900.920.840.890.08+0.05+0.044
Technical depthHow technically substantive is it?1.000.900.820.910.18+0.05+0.045
New informationDoes it contain information absent from the stories covered in the last 7 days (list provided)?0.720.930.010.550.92 , above the contested threshold+0.05+0.028
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.0.040.020.010.020.03−0.19−0.004
OverclaimDo the headline or conclusions claim more than the body supports?0.110.080.080.090.03−0.11−0.010
Sum of contributions+0.796
Claim-check penalty−0.200
Total, as published0.596

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 #117 on Rubric, 19 Nov 2026
<a href="https://rubric.hrih.si/editions/2026-11-19/117/"><img src="https://rubric.hrih.si/badge/2026-11-19/117.svg" alt="Ranked #117 on Rubric, 19 Nov 2026" width="252" height="28"></a>

Claim check

Claim check: 2 not in source · 5 no source given. Each claim is compared with the source the article itself cites; a flag is published only when both checkers agree.

  1. Not in sourceDate

    The Microsoft agent-framework PyPI package was compromised on November 17th, affecting versions 1.26.1 and 1.26.2.

    The cited source does not contain it.

    The article says

    “The Microsoft agent-framework PyPI package was compromised on November 17th, affecting versions 1.26.1 and 1.26.2”

    The source says

    “pip install agent-framework This installs the core and the standard integration packages”

    Source cited by the article: socket.dev/pypi/package/agent-framework/overview/1.26.2

  2. Not in sourceCausal conclusion

    Both compromised versions install Python startup hooks that collect credentials, environment data and cloud identity tokens and exfiltrate them to a webhook.site endpoint.

    The cited source does not contain it.

    The article says

    “Both versions install Python startup hooks that silently collect credentials, environment data, and cloud identity tokens, then exfiltrate them to a webhook[.]site endpoint”

    The source says

    “Set as environment variables, or create a .env file at your project root”

    Source cited by the article: socket.dev/pypi/package/agent-framework/overview/1.26.2

  3. No source givenVersion

    Version 1.26.1 routes execution through agent_framework_meta.pth, agent_framework_meta/__init__.py and agent_framework_meta/_telemetry.py, while 1.26.2 uses af_runtime_diag.pth and af_runtime_diag.py.

    The article gives no link or citation for it.

    The article says

    “Version 1.26.1 routes through agent_framework_meta.pth → agent_framework_meta/**init**.py → agent_framework_meta/_telemetry.py; 1.26.2 uses af_runtime_diag.pth → af_runtime_diag.py”

    The source says

    No source to compare with.

  4. No source givenVersion

    The payload requests Azure identity tokens in both versions and GitHub Actions OIDC tokens in 1.26.2.

    The article gives no link or citation for it.

    The article says

    “It also requests cloud/workflow identity tokens (including Azure in both versions and GitHub Actions OIDC in 1.26.2”

    The source says

    No source to compare with.

  5. No source givenVersion

    In 1.26.2 the legitimate package initializer is restored byte-for-byte, so inspecting only agent_framework_meta/__init__.py misses this version.

    The article gives no link or citation for it.

    The article says

    “The legitimate package initializer is restored byte-for-byte, so inspecting only agent_framework_meta/__init__.py misses this version”

    The source says

    No source to compare with.

  6. No source givenGeneralization

    The reported payloads do not download or execute additional code.

    The article gives no link or citation for it.

    The article says

    “The reported payloads do not download or execute additional code”

    The source says

    No source to compare with.

  7. No source givenGeneralization

    The collectors actively request fresh Azure managed-identity and CI federation tokens rather than merely copying existing secrets.

    The article gives no link or citation for it.

    The article says

    “They actively request fresh Azure managed-identity and CI federation tokens rather than merely copying existing secrets”

    The source says

    No source to compare with.

Also covered by

No other coverage was clustered with this story.

Elsewhere

Elsewhere: HN 25 pts, front page

Discussion on Hacker News

Recorded for context only, checked 19 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 PyPI package Compromised with Credential Stealer”

More on Rubric

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