supply-chain

Hunt LLM supply-chain compromise (OWASP LLM03:2025) — malicious or backdoored models, datasets, adapters, plugins, MCP servers, and tokenizer / framework dependencies that ship inside an AI-integrated product.

Quality

73%

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/standard/analyst/supply-chain/SKILL.md

Quality

Content

82%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.

A tight, highly actionable security-review skill body with executable inventory and PoC code and a solid validation contract. The main improvements are inline validation checkpoints in the audit workflow and moving some PoC detail into reference files for cleaner progressive disclosure.

Suggestions

Inline a brief validation checkpoint in the audit workflow (e.g. 'for each candidate artifact, run the validate_finding negative_command before reporting') so the feedback loop sits next to the steps rather than only in section 6.

Provide or generate the known-good.txt baseline referenced by the typo-squat PoC (or note that it must be supplied), so that snippet is copy-paste runnable end-to-end.

Consider splitting the PoC payloads (section 5) and the CVSS table (section 7) into a reference file so SKILL.md stays a lean overview with one-level-deep references.

Dimension	Reasoning	Score
Conciseness	The body is lean and purposeful — terse recognition-signal bullets, compact attack-vector descriptions, and minimal code blocks with no padding or explanation of concepts Claude already knows; nearly every token earns its place.	5 / 5
Actionability	Mostly executable guidance: copy-pasteable grep -rE inventory commands and a runnable Python trigger-sweep cover the common cases, with minor gaps such as the typo-squat script depending on an unprovided known-good.txt and one PoC being procedural rather than code.	4 / 5
Workflow Clarity	The audit workflow is a clear sequenced set of inventory steps plus a per-artifact checklist, and a validate_finding contract (success/negative patterns) supplies the validation feedback loop this batch operation needs; the only gap is that validation lives in a separate section rather than inline at each checkpoint.	4 / 5
Progressive Disclosure	The ~140-line body is well organized into eight clearly headed, one-level sections with a properly signaled cross-skill reference; with no bundle files present everything is inline, so structure is good but nothing is split into reference files.	4 / 5
	Total	17 / 20 Passed

Description

65%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.

A specific, distinctive description with strong trigger keywords, weakened mainly by a single action verb and the absence of an explicit 'Use when...' clause. Adding explicit activation guidance and one or two more concrete verbs would lift completeness and specificity.

Suggestions

Add an explicit 'Use when...' clause naming the trigger situations (e.g. 'Use when reviewing an LLM-integrated product for supply-chain compromise, or when the user mentions malicious models, backdoored adapters, or unpinned LLM framework dependencies').

Add one or two more concrete action verbs beyond 'Hunt' (e.g. 'inventory', 'fingerprint', 'validate') to broaden the specificity of capabilities.

Include common file extensions or artifact tokens (e.g. .gguf, .safetensors, LoRA) in the description to round out trigger-term coverage.

Dimension	Reasoning	Score
Specificity	The description names the domain comprehensively ('malicious or backdoored models, datasets, adapters, plugins, MCP servers, and tokenizer / framework dependencies') but offers only one concrete action verb ('Hunt'), so it names the domain and 1-2 actions without comprehensive action coverage.	3 / 5
Completeness	The 'what' is clear and concrete, but the description lacks any explicit 'Use when...' trigger clause; per the judging guidelines a missing explicit trigger guidance caps completeness at 3 even though a when is weakly implied by the OWASP reference.	3 / 5
Trigger Term Quality	Strong natural keyword coverage ('LLM supply-chain compromise', 'malicious', 'backdoored model', 'dataset', 'adapter', 'plugin', 'MCP', 'tokenizer', 'framework dependency', 'OWASP LLM03') that a reviewer would actually say, though a few natural variants and file extensions are missing.	4 / 5
Distinctiveness Conflict Risk	It carves a clear niche (LLM supply-chain compromise, OWASP LLM03) with distinct artifact-type triggers (models, LoRA adapters, MCP servers, tokenizers) and minimal overlap risk with other skills.	5 / 5
	Total	15 / 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/standard/analyst/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.