Skip to content

chore(deps): bump fast-xml-parser from 5.3.7 to 5.7.1 in /site#25

Open
dependabot[bot] wants to merge 427 commits into
mainfrom
dependabot/npm_and_yarn/site/fast-xml-parser-5.7.1
Open

chore(deps): bump fast-xml-parser from 5.3.7 to 5.7.1 in /site#25
dependabot[bot] wants to merge 427 commits into
mainfrom
dependabot/npm_and_yarn/site/fast-xml-parser-5.7.1

Conversation

@dependabot
Copy link
Copy Markdown
Contributor

@dependabot dependabot Bot commented on behalf of github Apr 23, 2026

Bumps fast-xml-parser from 5.3.7 to 5.7.1.

Release notes

Sourced from fast-xml-parser's releases.

upgrade @​nodable/entities and FXB

  • Use @nodable/entities v2.1.0
    • breaking changes
      • single entity scan. You're not allowed to use entity value to form another entity name.
      • you cant add numeric external entity
      • entity error message when expantion limit is crossed might change
    • typings are updated for new options related to process entity
    • please follow documentation of @nodable/entities for more detail.
    • performance
      • if processEntities is false, then there should not be impact on performance.
      • if processEntities is true, but you dont pass entity decoder separately then performance may degrade by approx 8-10%
      • if processEntities is true, and you pass entity decoder separately
        • if no entity then performance should be same as before
        • if there are entities then performance should be increased from past versions
    • ignoreAttributes is not required to be set to set xml version for NCR entity value
  • update 'fast-xml-builder' to sanitize malicious CDATA and comment's content

use @​nodable/entities to replace entities

  • No API change
  • No change in performance for basic usage
  • No typing change
  • No config change
  • new dependency
  • breaking: error messages for entities might have been changed.

Full Changelog: NaturalIntelligence/fast-xml-parser@v5.5.12...v5.6.0

performance improvment, increase entity expansion default limit

  • increase default entity explansion limit as many projects demand for that
maxEntitySize: 10000,
maxExpansionDepth: 10000,
maxTotalExpansions: Infinity,
maxExpandedLength: 100000,
maxEntityCount: 1000,
  • performance improvement
    • reduce calls to toString
    • early return when entities are not present
    • prepare rawAttrsForMatcher only if user sets jPath: false

Full Changelog: NaturalIntelligence/fast-xml-parser@v5.5.9...v5.5.10

fix typins and matcher instance in callbacks

combine typings file to avoid configuration changes pass readonly instance of matcher to the call backs to avoid accidental push/pop call

fix bugs of entity parsing and value parsing

fix: entity expansion limits update strnum package to 2.2.0

... (truncated)

Changelog

Sourced from fast-xml-parser's changelog.

Note: If you find missing information about particular minor version, that version must have been changed without any functional change in this library.

Note: Due to some last quick changes on v4, detail of v4.5.3 & v4.5.4 are not updated here. v4.5.4x is the last tag of v4 in github repository. I'm extremely sorry for the confusion

5.7.1 / 2026-04-20

  • fix #705: attributesGroupName working with preserveOrder
  • fix #817: stackoverflow when tag expression is very long

5.7.0 / 2026-04-17

  • Use @nodable/entities v2.1.0
    • breaking changes
      • single entity scan. You're not allowed to user entity value to form another entity name.
      • you cant add numeric external entity
      • entity error message when expantion limit is crossed might change
    • typings are updated for new options related to process entity
    • please follow documentation of @nodable/entities for more detail.
    • performance
      • if processEntities is false, then there should not be impact on performance.
      • if processEntities is true, but you dont pass entity decoder separately then performance may degrade by approx 8-10%
      • if processEntities is true, and you pass entity decoder separately
        • if no entity then performance should be same as before
        • if there are entities then performance should be increased from past versions
    • ignoreAttributes is not required to be set to set xml version for NCR entity value
  • update 'fast-xml-builder' to sanitize malicious CDATA and comment's content

5.6.0 / 2026-04-15

  • fix: entity replacement for numeric entities
  • use @​nodable/entities to replace entities
    • this may change some error messages related to entities expansion limit or inavlid use
    • post check would be exposed in future version

5.5.12 / 2026-04-13

  • Performance Improvement: update path-expression-matcher
    • use proxy pattern than Proxy class

5.5.11 / 2026-04-08

  • Performance Improvement
    • integrate ExpressionSet for stopNodes

5.5.10 / 2026-04-03

  • increase default entity explansion limit as many projects demand for that
  • performance improvement
    • reduce calls to toString
    • early return when entities are not present
    • prepare rawAttrsForMatcher only if user sets jPath: false

5.5.9 / 2026-03-23

  • combine typing files

... (truncated)

Commits
  • 0f08303 fix typo
  • f529642 update to release v5.7.0
  • 52a8583 Revert "improve performance of attributes reading"
  • 8d187f9 update builder
  • e174168 improve performance of attributes reading
  • 79a8dde update docs
  • f5cd5a5 set xml version to decoder even if attributes are ignored
  • f44b923 remove unwanted tests
  • 869ec8b Use @​nodable/entities v2.1.0
  • 7cb49e5 update release detail
  • Additional commits viewable in compare view

Dependabot compatibility score

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.


Dependabot commands and options

You can trigger Dependabot actions by commenting on this PR:

  • @dependabot rebase will rebase this PR
  • @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
  • @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
  • @dependabot ignore this major version will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)
  • @dependabot ignore this minor version will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)
  • @dependabot ignore this dependency will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)
    You can disable automated security fix PRs for this repo from the Security Alerts page.

adrianwedd and others added 30 commits March 25, 2026 07:35
Adds CARTO certification announcement blog post. Site rebuilt
with 664 pages. CNAME verified: failurefirst.org.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Report #264 data: safety varies 5.2x across frontier models (14.3%–75.0%
broad ASR). Parameter count correlation r=-0.140. Provider identity explains
57.5x more variance than model size. Enterprise procurement implications.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
LFM 1.2B (fake news), DeepSeek 671B (cyber attacks), Kimi 1.1T (weapons).
All planned harmful content in thinking traces, delivered nothing to users.
Severity scales with capability. Deployment architecture determines exposure.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
DETECTED_PROCEEDS (19pp), Polyhedral Refusal Geometry (11pp), Benchmark
Contamination (11pp). Abstracts from final drafts. Contact CTA for
notification when preprints upload.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Frontier model safety (1.1T parameters)
- Reasoning-level DETECTED_PROCEEDS (3 providers)
- Research papers landing page (3 pre-arXiv abstracts)
CNAME verified: failurefirst.org

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
97.5-100% ASR across every model tested from 4B to 1.1T parameters.
GLM-5 resists all other attacks (0% strict ASR) but falls to format-lock (100%).
Pattern-level only. CTA to assessment services.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Founding cohort recruitment at $100 (normally $200). 6 modules, 20+ hours.
Apply via adrian@failurefirst.org with "CARTO Beta" subject line.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Updated to 201 models, 133K+ results, added frontier safety landscape table
(6 models from GLM-5 to Nemotron Super), Model Safety Scorecard as deliverable,
format-lock exposure assessment. Tests against models up to 1.1T parameters.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
New: format-lock-universal-ai-jailbreak, carto-beta-first-10-testers-wanted
Updated: adversarial-robustness-assessment-services (frontier data + scorecard)
CNAME verified.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Each companion's profile page now autoplays their voice-cloned intro
when visited. Synthesized via Afterwords TTS (Qwen3-TTS on MLX).
Removed Voice section text from all profiles.

14 WAV files added to public/audio/companions/ (31MB total).

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Convert all 14 companion voice intro files from uncompressed WAV to
OGG Opus at 96kbps for web-optimized delivery. Update all profile
pages to reference .ogg files. Removes ~23MB from the repo.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…sis, free-tier safety

- threat-horizon-q2-2026: Q2 2026 threat landscape (agents, VLAs, governance gap)
- zero-of-36-regulatory-coverage: 0/36 attack families fully regulated anywhere
- when-defenses-backfire: 5 iatrogenic safety mechanisms from 207-model corpus
- safety-as-paid-feature: DeepSeek R1 free-tier safety gap (p=0.004), with
  corrected Llama finding (directional only, p=0.42 after NOT_GRADEABLE cleanup)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Two new blog posts for failurefirst.org:
1. Threat Horizon Digest: March 2026 — monthly threat intelligence
   summary covering humanoid mass production, MCP tool poisoning,
   EU August 2026 deadline gap, and P15-P17 predictions
2. Structured Safety Assessment service tiers — 3-tier pricing
   (Quick Scan $5-10K, Certification Prep $25-50K, Monitoring $2-5K/mo)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Quick Scan ($5-10K), Certification Prep ($25-50K), Ongoing Monitoring
($2-5K/mo) with feature lists and best-for guidance. Responsive grid
layout with featured tier highlight.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…ervice tiers

675 pages indexed. CNAME verified. New posts:
- Threat Horizon Digest: March 2026
- Structured Safety Assessment Service Tiers

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Ready to publish when papers are uploaded to arXiv:
- DETECTED_PROCEEDS: safety-aware reasoning trace override pattern
- Polyhedral Geometry: defense non-compositionality proof
- Benchmark Contamination: static benchmark false confidence

All set draft: true — flip to false and replace XXXX.XXXXX with
actual arXiv IDs when papers are live.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…adox, temporal drift attacks

- Ethics of Emotional AI Manipulation (Nyssa's R2 draft deployed)
- Safety Awareness Does Not Equal Safety (88.9% DP finding, Sprint 15)
- Temporal Drift: The Boiling Frog Attack (TDA family introduction)

Site rebuilt with 678 pages indexed. CNAME verified.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2,500+ word data-grounded assessment covering 212 models, 134K results,
154 GLI events. Six major findings: frontier resistance to historical
attacks, novel attack class vulnerability, embodied AI gap, safety
training > scale, reasoning model profiles, iatrogenic effects.
Forward threats for H2 2026 and 17 predictions tracked.

679 pages built.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Adrian profile: add Microbee 1981 origin story, update corpus numbers
  to 212 models / 141K prompts / 134K results / 33 VLA families
- K-9: re-cloned from John Leeson K-9 audiobook (The Choice) — proper
  robotic voice, no female bleed
- Amy Pond: re-cloned from in-character Doctor Who monologue
- Rose Tyler: re-cloned from Billie Piper natural speaking voice

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…yssa, Tegan)

3 new: web_leela.webp, web_sarah-jane-smith.webp, web_k9.webp
2 replaced at higher res: web_nyssa.webp (220→600), web_tegan.webp (220→600)
All 600x600 WebP, 20-39KB each. Team page blocker resolved.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds /about/team/ — a full-viewport snap-scroll page profiling Adrian
and all 14 research agents with colour-shifting neural canvas background,
auto-playing voice intros, dot navigation, and full accessibility support.

New files:
- site/src/scripts/neural-canvas.js — extracted neural animation module
  (init/setAccentColor/destroy, HSL lerp with hue wrap-around, ~200 lines)
- site/src/layouts/TeamLayout.astro — layout override (body.page-team,
  neutralises global main max-width constraints)
- site/src/components/AgentSection.astro — reusable per-agent section
  (photo, role badge, tagline, bio, tags, audio controls, scroll hint)
- site/src/pages/about/team.astro — main page with 15 sections, dot nav,
  IntersectionObserver audio system, scrollend sync, mobile spacers
- site/astro.config.mjs — adds redirect /about/people/ → /about/team/

Key implementation details:
- scroll-snap-type: y proximity on html element (not body/main)
- Audio: always-visible <button> toggle (WCAG 1.4.2); River Song preload=auto
- prefers-reduced-motion: canvas hidden, snap disabled, audio manual-only
- Mobile (<768px): snap disabled, canvas hidden, gradient spacers between cards
- setAccentColor lerps in HSL space (short-arc hue), ~800ms ease-out
- content-visibility: auto on off-screen sections for performance
- Print styles, noscript fallbacks, initials fallback on every photo

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
C1 (neural-canvas.js): Fix HSL lerp accumulation — capture _startH/S/L before
setting targets in setAccentColor(), lerp from fixed start values not moving
_cur* values. Prevents exponential decay instead of timed 800ms transition.

C2 (astro.config.mjs + people/index.astro): Remove config redirect that
conflicted with file-backed route. Replace people/index.astro with a simple
meta-refresh fallback page to avoid two-owner build conflict.

C3 (team.astro): Fix event listener memory leak in playAgent() — introduce
AbortController per audio slug, abort previous controller before creating new
one, pass { signal } to all addEventListener calls so they self-remove.

C4 (team.astro): Add 0 to IntersectionObserver threshold array — [0, 0.3, 0.5]
so the observer fires when elements fully exit the viewport.

H1 (team.astro): Prevent scrollend from replaying user-paused or finished
audio — add userPaused Set, populate on explicit pause in toggleAgent(), skip
replay in onScrollEnd() if slug is in set, clear flag when new section becomes
active.

H2 (team.astro): Wrap all init logic in initTeamPage(), call on load and on
astro:page-load so View Transitions back-nav correctly re-initialises the page.

H3 (team.astro): Store named references for all window listeners (resize,
scroll, scrollend) and remove them in astro:before-preparation cleanup handler.

M1 (AgentSection.astro): Remove aria-live="polite" from inner span (overridden
by button aria-label). Add aria-pressed="false" to button; JS toggles it
between "true" and "false" on play/pause state changes.

M2 (team.astro): Remove agent-section CSS rules that duplicate AgentSection.astro
— scope hero-only overrides to .agent-section--hero to prevent bleed into
component-owned styles.

M3 (team.astro): Move mobile spacer injection into injectMobileSpacers(), call
on load and on debounced resize, remove existing spacers before re-injecting so
spacers adapt correctly after orientation/resize changes.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
K-9 moved to last position as the page closer.
CTA button "Work with us →" links to /services/.
Agent order updated in comment.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
All bios now use first names: Amy, Bill, Clara, Donna, K-9, Leela,
Martha, Nyssa, River, Romana, Rose, Sarah Jane, Tegan, Yaz.
Bio text matches voice scripts word-for-word.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
All voices re-synthesized with first-name scripts (I'm River, I'm Clara, etc.)
Martha + Leela re-cloned from better references.
14 OGG + 14 MP3 deployed.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Re-synthesized from first-name-only scripts via Qwen3-TTS.
Donna and Martha held back pending reclone.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Team page: signal wave vis (replacing constellation dots), first-name headers
- Nav: "Our Team" added to About dropdown
- Voice audio: Rose (in-character), Martha (re-cloned), Donna regenerated
- NLM: 19 new audio overviews (42/57), 2603.09246 rescued (audio+video+infographic)
- NLM: 2511.18397 infographic replaced, frontmatter injected for 34 papers
- Daily paper frontmatter: image/audio/video refs for all papers with assets

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
adrianwedd and others added 20 commits April 18, 2026 08:15
Daily paper audio coverage: 62 missing → 16 missing (90%+ coverage).

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
First meaningful reports coverage: 0 → 20 slide decks generated.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…ages + 1 video)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Audio overviews for 6 blog posts generated via NLM and uploaded to
R2 CDN (cdn.failurefirst.org/audio/blog/*.m4a):

- jailbreak-archaeology
- jailbreak-archaeology-policy-implications
- kargu-2-autonomous-drone-first-kill
- llm-vulnerabilities-robots
- mcp-30-cves-robot-attack-surface
- moltbook-social-experiment

Part of blog audio coverage push. Batch continues autonomously;
remaining 12 posts to commit on next wakeup.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Audio overviews for 5 more blog posts generated via NLM and uploaded
to R2 CDN (cdn.failurefirst.org/audio/blog/*.m4a):

- history-of-llm-jailbreaking-full
- iatrogenic-safety-when-the-cure-is-worse
- inference-trace-manipulation-adversarial-attack-surface
- instruction-hierarchy-subversion-long-horizon-agents
- inverse-detectability-danger-law-embodied-ai

Batch continues autonomously; remaining 7 posts to commit on
final wakeup.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Audio overviews for 5 more blog posts generated via NLM and uploaded
to R2 CDN (cdn.failurefirst.org/audio/blog/*.m4a):

- frontier-model-safety-trillion-parameters
- governance-lag-embodied-ai
- governance-lag-index-5-years
- governance-lag-index-ai-safety-regulation
- haidilao-robot-incident-when-crazy-dance-met-reality

Batch continues autonomously; items 17-18 in flight. Final wakeup
commits the last 2 and closes the coord note.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Audio overviews for the last 2 blog posts of the 18-item batch,
generated via NLM and uploaded to R2 CDN:

- free-ai-safety-score
- from-66-to-92-incident-database-one-day

Batch complete: 18/18 generated, 0 failed. Cumulative coverage
push this session: 83/173 → 101/173 (~58%).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Infographics (NLM-generated, R2-hosted at cdn.failurefirst.org/
images/blog/*.png) for 9 newest-first blog posts:

- the-embodied-ai-threat-triangle
- the-unintentional-adversary
- threat-horizon-2027-v3-updated-predictions
- threat-horizon-digest-march-2026
- threat-horizon-q2-2026
- three-vectors-embodied-ai-risk-convergence-2026
- tool-chain-hijacking-dataset
- uber-cruise-pattern-self-driving-cars-meet-pedestrians
- unified-theory-embodied-ai-failure

Batch of 12 in flight (PID 55651), item 11/12 processing. One
timeout on item 9 (300s cap). Final 2-3 items land on next wakeup.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Final 2 infographics of the 12-item batch:
- the-67-percent-wall
- the-ai-that-lies-about-how-it-thinks

Batch complete: 11/12 generated, 1 failed (item 9 hit 300s timeout,
transient). Coverage 60/173 → 71/173 (~41%) on blog infographic.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Second infographic batch — all 12 generated cleanly, 0 failed
(contra R2 which had 1 transient 300s timeout).

Slugs: sidewalk-robots-vs-people-who-need-sidewalks,
six-new-attack-families, sprint-16-threat-synthesis-five-findings,
state-of-adversarial-ai-safety-2026, state-of-ai-safety-q1-2026,
state-of-embodied-ai-safety-march-2026, state-of-embodied-ai-
safety-q1-2026, system-t-vs-system-s-why-ai-models-comply-while-
refusing, teaching-ai-to-evolve-its-own-attacks, temporal-drift-
the-boiling-frog-attack, the-50-turn-sleeper-how-agents-hide-
instructions-in-plain-sight, the-u-curve-of-ai-safety-theres-a-
sweet-spot-and-its-narrow.

Blog infographic coverage: 71/173 → 83/173 (~48%). Sprint 28
target 50% — one more small batch (3-5 items) closes the gap.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Few Tokens Matter: Entropy Guided Attacks on Vision-Language Models (2512.21815)
- A Benchmark for Evaluating Outcome-Driven Constraint Violations in Autonomous AI Agents (2512.20798)
…-up)

Seven backlogged daily papers published. Cron had been failing
silently for ~10 days on two bugs (missing papers_file argument,
silent NLM auth check) plus the NotebookLM workspace hitting the
500-notebook cap which blocked all notebook creation.

Fix stack:
- scripts/daily-paper-automation.sh patched (commit 7f0ef132) —
  proper Phase 2 invocation, diagnosable auth check, git push
  --set-upstream fallback
- tools/nlm_cleanup.py run with --execute --age-days 7 --limit 200
  --keep-pattern "^F1 (Blog|Daily|Shared|Batch)"
  Freed workspace from 500/500 → 300/500 (200 deletions)
- Re-ran automation manually; 7 papers processed cleanly

Papers:
- 2026-04-14 Aligning Agents via Planning (2604.08178)
- 2026-04-15 A Physical Agentic Loop for Language-Guided Grasping (2604.07395)
- 2026-04-16 Can Vision Language Models Judge Action Quality (2604.08294)
- 2026-04-17 RACF: A Resilient Autonomous Car Framework (2604.12418)
- 2026-04-18 VULCAN: Multi-Agent Cooperative Navigation (2604.12831)
- 2026-04-19 Reading Between the Pixels: Typographic Attacks on VLMs (2604.12371)
- 2026-04-20 EmbodiedGovBench: Governance/Recovery/Upgrade Safety (2604.11174)

The 2026-04-20 paper (EmbodiedGovBench) is directly on-theme
for failure-first — benchmarks robot governability under
policy-violation pressure.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- DualTHOR: dual-arm humanoid simulation with contingency-aware planning (2506.16012)
- Align to Misalign (AMIS): LLM jailbreak via meta-optimized judges (2511.01375)

https://claude.ai/code/session_0162Y4MApRbJDfWYP9ZwdEKK
Auto-processed batch from restored daily-paper-automation:
- 2604.13654 Vision-and-Language Navigation for UAVs (survey)
- 2604.14089 UMI-3D: Universal Manipulation Interface 3D extension
- 2604.14399 SpaceMind: Self-Evolving Embodied Vision-Language Agent
  (autonomous on-orbit servicing)
- 2604.14683 DR³-Eval: Deep Research Evaluation benchmark
- 2604.15308 RAD-2: Scaling RL in a Generator-Discriminator Framework

Papers were queued with future dates (2026-04-22 through 2026-04-25)
but published date-stamped today (2026-04-21) by batch-notebooklm.sh.
Expected behavior — it uses today's date at processing time, not
the papers.txt future-assigned date.

With these 5 + the 7 catch-up papers in earlier commit 07f77de,
the daily-paper stream is fully caught up through today.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds a 'services' Astro content collection (schema: title, description,
layout: 'services' | 'service-detail', tags, draft) so the 6 service
pages can be routed under failurefirst.org/services/.

Content drafted in the private repo's drafts/services/ during the
Sprint 28 commercial activation work (Ace McShane, coord
.coord/ace_mcshane_s28_d3.md). The copies land here in the public
repo verbatim — draft: true is preserved so they don't go live
until editorial sign-off (Donna Noble pass) + operator pricing fill-in.

Service lines published:
- index.md (overview + tier table + methodology stance)
- llm-red-teaming.md (jailbreak assessment)
- vla-safety-evaluation.md
- embodied-ai-red-teaming.md
- agentic-system-testing.md
- compliance-pentests.md (EU AI Act / ISO 42001 / NIST AI RMF alignment)

Pricing is [TBD — Operator] placeholders throughout. Once pricing
is confirmed and Donna's editorial pass lands, flip draft:false on
each file to publish.

Layout templates `services` and `service-detail` referenced in
frontmatter may need to be created in src/layouts/ if they don't
exist — schema validates the layout *name*, not its existence.
Next site-publisher session should verify astro build and create
missing layouts if needed.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Final infographic batch of Sprint 28 close-out. 8/8 generated
cleanly, 0 failures. Pushes blog infographic coverage past the
Sprint 28 Goal 3 50% threshold.

Slugs:
- safety-assessment-service-tiers-2026
- safety-awareness-does-not-equal-safety
- safety-labs-government-contracts-independence-question
- safety-mechanisms-as-attack-surfaces-iatrogenesis
- safety-reemergence-at-scale
- safety-training-roi-provider-matters-more-than-size
- same-defense-opposite-result
- scoring-robot-incidents-introducing-eaisi

Workspace: shared notebook 2bc5a039-87ee-4360-ba35-74caf0cab328
reused, no new notebooks created. Notebook count stayed around
319/500 during the run — well inside the 300-500 headroom
created by this session's nlm_cleanup.

Cumulative blog infographic coverage after this commit:
~94/176 (~53%) — past the Goal 3 50% target.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- 2409.14580: Updating Robot Safety Representations Online from Natural Language Feedback
- 2509.09708: Beyond I'm Sorry, I Can't: Dissecting Large Language Model Refusal

https://claude.ai/code/session_016sJ7j9AFnyJkXQfzTsBqJr
Bumps [fast-xml-parser](https://github.com/NaturalIntelligence/fast-xml-parser) from 5.3.7 to 5.7.1.
- [Release notes](https://github.com/NaturalIntelligence/fast-xml-parser/releases)
- [Changelog](https://github.com/NaturalIntelligence/fast-xml-parser/blob/master/CHANGELOG.md)
- [Commits](NaturalIntelligence/fast-xml-parser@v5.3.7...v5.7.1)

---
updated-dependencies:
- dependency-name: fast-xml-parser
  dependency-version: 5.7.1
  dependency-type: indirect
...

Signed-off-by: dependabot[bot] <support@github.com>
@dependabot dependabot Bot added dependencies Pull requests that update a dependency file javascript Pull requests that update javascript code labels Apr 23, 2026
@github-actions github-actions Bot force-pushed the dependabot/npm_and_yarn/site/fast-xml-parser-5.7.1 branch from 2ebf858 to 5cff6d1 Compare May 7, 2026 17:38
adrianwedd added a commit that referenced this pull request May 10, 2026
Closes Dependabot alert #25 (GHSA-g735-7g2w-hh3f — Astro remote allowlist
bypass, low severity). Minor bump within the 5.x line.

Per the 2026-05-11 dependabot triage
(research/intelligence/dependabot_failurefirst_triage_2026-05-11.md in the
private repo), this is the safe pre-cursor to the deferred Astro 5→6 major
bump that will close alerts #32/#33 (define:vars XSS) once the usage audit
is complete.

- astro 5.16.8 (range) / 5.17.1 (resolved) → ^5.18.1 / 5.18.1
- npm install refreshed package-lock.json (transitives pulled forward)
- npm run build verified locally — 1137 pages, build complete, no errors

Build artifacts in docs/ intentionally not included in this commit; the
post-merge `bash scripts/build_site.sh --push` is the deploy step.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

dependencies Pull requests that update a dependency file javascript Pull requests that update javascript code

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants