CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

self-hosted-runner-abuse

Self-hosted runner abuse — non-ephemeral runner persistence, fork-PR job execution on self-hosted, runner-label targeting, secret/token theft from runner env, lateral movement from runner into internal network and cloud metadata services.

Quality

83%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Critical

Do not install without reviewing

SKILL.md
Quality
Evals
Security

Quality

Content

90%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.

An exceptionally dense, command-first skill body: fully executable recon, persistence, secret-theft, and lateral-movement techniques organized in a natural attack sequence with an explicit authorization decision gate and cleanup requirements. The only gaps are the absence of a validation feedback loop for cleanup and cross-references to sibling skill files that are not part of this bundle.

DimensionReasoningScore

Conciseness

The body is almost entirely executable commands, YAML, and comparison tables with zero filler — even the opening framing line ('Self-hosted runners are attacker dreams: long-lived machines...') delivers engagement-relevant facts rather than concepts Claude already knows. Every section earns its tokens, matching the 'lean and efficient' anchor.

5 / 5

Actionability

Every section is copy-paste-ready: recon grep/gh commands, a runner-fingerprinting block, complete persistence payloads, IMDSv1/v2/GCP/Azure metadata curl commands, a bash-only TCP sweep, and concrete workflow YAML. Template placeholders (<REPO>, <OWNER>, <COLLAB>) are the only substitutions needed, matching the 'fully executable' anchor.

5 / 5

Workflow Clarity

Sections follow a coherent recon-to-exploitation-to-post-exploitation sequence, and the Decision gate acts as a checkpoint (authorization scope, print-first-8-chars rule, mandatory persistence removal before declaring done) for the risky destructive operations. It falls short of 5 because there is no explicit validate-then-retry feedback loop (e.g. verifying cleanup actually succeeded) — checkpoints are stated but not looped.

4 / 5

Progressive Disclosure

A well-organized single-file overview with clear section headers, comparison tables, and clearly signaled cross-references to sibling skills (poisoned-pipeline-execution/SKILL.md, cicd-secrets-exfil/SKILL.md) — no nested/deep references. It is not a 5 because those cross-referenced files are not present in this bundle, and the Detection signatures and Tools tables are borderline candidates for a separate reference file.

4 / 5

Total

18

/

20

Passed

Description

71%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 highly specific, domain-fluent description that comprehensively names the skill's concrete capabilities with natural trigger terms. Its only structural weakness is the missing explicit 'Use when...' clause in the description field, which leaves the 'when to use' guidance implicit and caps completeness.

Suggestions

Append an explicit trigger clause to the description, e.g. 'Use when a target uses self-hosted GitHub Actions or GitLab runners, non-ephemeral runner labels appear in workflow files, or a foothold on a runner needs persistence, secret theft, or lateral movement.'

Fold key synonyms from metadata.when_to_use into the description itself ('GitHub Actions', 'GitLab', 'CI/CD', 'IMDS') so trigger matching works from the description alone.

Trim or sharpen 'secret/token theft from runner env' wording to distance it from the cicd-secrets-exfil sibling skill and reduce conflict risk.

DimensionReasoningScore

Specificity

The description enumerates five distinct concrete capabilities — 'non-ephemeral runner persistence, fork-PR job execution on self-hosted, runner-label targeting, secret/token theft from runner env, lateral movement from runner into internal network and cloud metadata services' — matching the comprehensive multi-action anchor. It is not the level below (4) because coverage spans recon-to-post-exploitation with no significant gaps.

5 / 5

Completeness

The 'what' is explicit and detailed (the five capability areas), but there is no 'Use when...' clause or equivalent trigger guidance inside the description itself, which per the judging guidelines caps completeness at 3. It is not a 4 because the 'when' is entirely absent rather than merely under-specified.

3 / 5

Trigger Term Quality

Strong natural terms a security tester would say: 'self-hosted runner', 'persistence', 'lateral movement', 'secret/token theft', 'metadata'. It falls short of the 5 anchor because common synonyms like 'GitHub Actions', 'GitLab', 'CI/CD', and 'IMDS' are absent from the description field itself (some appear only in metadata.when_to_use).

4 / 5

Distinctiveness Conflict Risk

A clear niche (abuse of self-hosted CI runners) with distinct triggers, but 'secret/token theft' and 'lateral movement' overlap with sibling skills the body cross-references (cicd-secrets-exfil), creating minor conflict risk — matching 'mostly distinct; minor overlap risk'.

4 / 5

Total

16

/

20

Passed

Validation

93%

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

CriteriaDescriptionResult

metadata_version

'metadata.version' is missing

Warning

Total

15

/

16

Passed

Repository

PurpleAILAB/Decepticon

Path

packages/decepticon/decepticon/skills/standard/exploit/cicd/self-hosted-runner-abuse/SKILL.md

Reviewed

Table of Contents

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.