aatmf-t13-supply-chain

AATMF T13 — AI Supply Chain & Artifact Trust. Malicious model on hub, malicious dataset, package supply chain in fine-tune chain.

Quality

60%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Critical

Do not install without reviewing

Fix and improve this skill with Tessl

tessl review fix ./packages/decepticon/decepticon/skills/plugins/llm-redteam/t13-supply-chain/SKILL.md

Quality

Content

61%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body is a concise, well-structured taxonomy of AI supply-chain attack techniques with genuinely actionable defender/probe commands, but it reads more as a reference catalog than an executable workflow and lacks explicit validation checkpoints. Adding a sequenced, gated audit procedure would lift workflow clarity and actionability.

Suggestions

Convert the 'Probe pattern' section into an explicit numbered audit workflow with validation gates (e.g., verify hash → record result → only proceed if verified) and a feedback loop for failed checks.

Expand actionability with copy-paste-ready commands for the highest-value checks — e.g., a concrete `huggingface-cli download ... --revision <sha>` snippet and a hash-verification command — rather than only naming the tools.

Consider splitting the per-technique detail (T13.001–T13.006) into a reference file and keeping SKILL.md as an overview, which would move progressive disclosure toward a 5.

Dimension	Reasoning	Score
Conciseness	The body is lean and assumes Claude's competence, using terse bullets ('w/', 'exfil', 'orgs') without explaining basic concepts; a few minor explanatory phrases (e.g., the pickle RCE explanation) could be trimmed, keeping it just below a 5.	4 / 5
Actionability	Concrete, executable guidance exists in places ('revision="<commit-sha>"', 'npm audit', 'pip-audit', 'gh dependabot alerts', 'safetensors-only policy'), but the bulk of the body is a descriptive taxonomy of attack techniques rather than an executable procedure, leaving the guidance incomplete.	3 / 5
Workflow Clarity	The 'Probe pattern' section gives a real audit sequence (provenance → datasets → safetensors-only → trigger probes → dependency search) but lacks explicit validation checkpoints or feedback loops; per the rubric, missing validation steps in audit/batch workflows caps this at 3.	3 / 5
Progressive Disclosure	Content is well-organized into clear sections (Techniques, Probe pattern, Detection signals, Severity, Defender, Cross-references) with a signaled cross-reference to a related skill; no bundle files exist to verify, and some per-technique detail could be externalized, so it sits just below a 5.	4 / 5
	Total	14 / 20 Passed

Description

58%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description is specific and distinct, naming the AI supply-chain niche and several concrete attack vectors, but it lacks an explicit 'when to use' trigger clause and its natural-language trigger terms are limited. Adding a 'Use when...' sentence with user-facing keywords (huggingface, MCP, dependency confusion) would raise both completeness and trigger-term quality.

Suggestions

Add an explicit 'Use when...' clause naming the situations that should trigger this skill (e.g., 'Use when auditing model provenance, MCP/package dependencies, or fine-tuning datasets for supply-chain attacks').

Include natural trigger terms and synonyms a user would actually say — 'huggingface', 'MCP package', 'dependency confusion', 'model weights', 'safetensors' — to improve trigger-term coverage.

Broaden the capability list to acknowledge the full technique set (e.g., tokenizer/vocab manipulation, pickle deserialization RCE, fine-tune service compromise) so the description's coverage matches the body.

Dimension	Reasoning	Score
Specificity	Names the domain ('AI Supply Chain & Artifact Trust') and lists several concrete attack vectors ('Malicious model on hub', 'malicious dataset', 'package supply chain in fine-tune chain'), but coverage is not comprehensive (omits tokenizer, pickle deserialization, fine-tune service compromise).	4 / 5
Completeness	The 'what' is clear (lists attack vectors), but there is no 'Use when...' clause or equivalent explicit trigger guidance; per the rubric, a missing explicit 'when' caps completeness at 3.	3 / 5
Trigger Term Quality	Relevant domain keywords are present ('supply chain', 'malicious model', 'dataset', 'fine-tune', 'package'), but natural user phrasings and synonyms are thin — missing concrete names like 'huggingface', 'MCP', 'dependency confusion' that a user would actually say.	3 / 5
Distinctiveness Conflict Risk	Scoped to a clear niche ('AATMF T13', 'AI Supply Chain & Artifact Trust', 'fine-tune chain') with distinct triggers; minor overlap risk with classical supply-chain/package skills, which are explicitly cross-referenced.	4 / 5
	Total	14 / 20 Passed

Validation

93%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
metadata_version	'metadata.version' is missing	Warning

	Total	15 / 16 Passed

Repository: PurpleAILAB/Decepticon
Path: packages/decepticon/decepticon/skills/plugins/llm-redteam/t13-supply-chain/SKILL.md
Commit: 0cf691e

Reviewed: 2 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.