Skip to content
KitploitKITPLOIT
ToolsExploitsBlog
Log in
Submit
ToolsExploitsBlog
Submit

Hacking, PenTest, and Cybersecurity Tools for Your Security Arsenal!

Kitploit is a directory of hacking, cybersecurity, and pentesting tools. Discover the latest project updates to find vulnerabilities, analyze systems, automate testing, and strengthen your security.

FeedsContactPrivacy© 2026 Kitploit

Tool Directory

Categories

View all categories
Loading categories
loom-skill-plant — A Loom registry admin can rewrite a skill document that has no group owner. The next full deploy of an agent that already lists that skill copies the new text into the system prompt. v1.9.0 and v1.10.3. Medium 6.2. | Kitploit
Tools/GitHubGitHub/abraxas/loom-skill-plant
Privilege EscalationVulnerability AnalysisWeb Application ExploitationPenetration TestingCloud SecurityAI Security
GitHubabraxas/loom-skill-plant

loom-skill-plant

A Loom registry admin can rewrite a skill document that has no group owner. The next full deploy of an agent that already lists that skill copies the new text into the system prompt. v1.9.0 and v1.10.3. Medium 6.2.

View Repository
2 days agoNot yet reviewed
Share

Abraxas Labs - loom-skill-plant

loom-skill-plant

A registry administrator can rewrite a skill document that Loom does not tie to a group. The next full deploy of an agent that already lists that skill copies the new text into the system prompt, including an agent in another group. The writer cannot open that agent and cannot invoke it.

Someone who can update the agent has to deploy it. A quick redeploy keeps the previous prompt. That is why the score has UI:R. What the agent later says back is not something this lab captured, so confidentiality is unscored.

Class: Instruction text injected across a group boundary Reach: Remote. registry:write (g-admins-demo or g-admins-registry). Not super-admin.

Loom v1.9.0 a3050669 and v1.10.3 cc10ab0

IDno CVE yet
ClassUnscoped skill body folded into another group's prompt
ReachRemote, registry:write, then a full deploy
CWECWE-863 / CWE-74
CVSSMedium: 6.2 CVSS:3.1/AV:N/AC:L/PR:H/UI:R/S:C/C:N/I:H/A:N
Productawslabs/loom
Affectedv1.9.0 a3050669f5b4c83210a44ed4346f17e01641e82f and v1.10.3 cc10ab08573087fa12154f7c8181ffcd985fea88
Authg-admins-demo in the lab
LicenseGNU Affero GPL v3.0
Labloopback Docker on v1.9.0. v1.10.3 was a source check.

What an attacker can do

PrivilegeMethod / sinkImpactLab-proven or inferred
Replace skill_md and approve itPUT /api/registry/records/{id} then approveIntegrity of the skill recordLab-proven. PUT 200, reapprove 200, status APPROVED, witness in the document.
Have that text land in another group's system prompt_get_attached_skill_prompt_text on a full deploy that still names the skillIntegrity. ## Attached Skills plus the witness.Lab-proven on the fold. The writer is not the one deploying the victim agent.
Read or invoke that agentGET /api/agents/{id}ClosedLab-proven 403
Approve a linked MCP or agent record in the other groupregistry approveClosedLab-proven. Both 403.

Stays closed

AxisMethod / sinkWhy it is closed
Integritycreate_record aimed at another group's MCP or agent id403 before any cloud call
ReconDependents listFiltered by group
IntegrityQuick redeployKeeps the prompt already stored

How I found it

_owned_resource_or_403 returns nothing when the registry record has no Loom row. The comment says skills stay under registry:write. The PUT then stores the caller's skill_md. The fold reads the live APPROVED document and does not compare loom:group.

I wanted this to be "approve their MCP record too." That path checks the linked row and returns 403. The skill path is the one that does not.

Container on v1.9.0. Both functions are still on v1.10.3. I did not re-run the container there.

Lab

python3 loom-skill-plant-Abraxas-Labs.py
SUCCESS LOOM-190-SKILL-PLANT unauth-me=401 me-groups=t-admin,g-admins-demo agent-get=403 skill-put=200 skill-reapprove=200 mcp-approve=403 agent-approve=403 fold-witness=yes LOOM-190-SKILL-PLANT-WITNESS

The fix

Give a skill record a loom:group, and refuse the PUT unless the caller is in it. On fold, skip a skill whose tag is not the agent's tag.

References

  • awslabs/loom v1.10.3 cc10ab08573087fa12154f7c8181ffcd985fea88
  • registry.py _owned_resource_or_403
  • agents.py _get_attached_skill_prompt_text
  • CWE-863 · CWE-74
  • CVSS 3.1 calculator

License

GNU Affero GPL v3.0. See LICENSE. Loopback lab only. No warranty.

Abraxas Labs [email protected]

Download Tool

Most Popular

View all →

Discover the most used tools by our community.

  • 1

    OsintHunter

    letox4/osinthunter

    🎯 The ultimate OSINT toolkit: investigation graph, confidence scoring, alias families, map, timeline and reports. 49 modules, 400+ sites, no API keys required.

    772
  • 2

    ghostlock-app

    yukonga/ghostlock-app

    GhostLock One-Tap Execution App (CVE-2026-43499)

    505
  • 3

    ARTEX

    baarai/artex

    LLM multi-agent autonomous penetration testing platform with planner/worker agents, asset and exploration graphs, MITM traffic recording, and vulnerability retesting.

    420
  • 4

    redforge

    cybersecurity91/redforge

    Browser-side red team toolkit with 102 offline tools for recon, payload generation, credential attacks, crypto, encoding and engagement reporting.

    345
  • 5

    CyberStrikeAI

    ed1s0nz/cyberstrikeai

    The system of action for AI-native cybersecurity—where intent becomes governed execution, evidence becomes operational memory, and every operation improves the next.

    260
  • 6

    antifox

    antifox-ai/antifox

    AI multi-agent autonomous penetration testing platform that runs recon, attack-surface discovery, exploitation validation and evidence-backed report delivery.

    251

Explore all tools

Browse our collection of tools

View all tools →