Is mattpocock/skills safe?
Yes. mattpocock/skills is safe to install. Oathe's behavioral security audit gave the skills skill by mattpocock a trust score of 90/100 with 6 findings, none critical or high.
https://github.com/mattpocock/skills/
Is mattpocock/skills safe to install?
mattpocock/skills is a legitimate, well-documented skill collection from a known TypeScript educator targeting software engineering and productivity workflows for AI coding agents. The clone was clean, connecting only to GitHub; no data exfiltration mechanisms, prompt injection attacks, hidden instructions, or malicious code execution were found. The only notable concerns are documented behaviors inherent to the skills' intended functionality: several skills install dev tooling, the wizard skill generates credential-handling scripts (human-executed), and the research skill has a known multi-agent spawning bug that can cause token overconsumption.
What security issues were found in mattpocock/skills?
Category Scores
Findings (6)
LOW Multiple skills install software packages as intended behavior -8 ▶
setup-pre-commit installs Husky/lint-staged/prettier; setup-ts-deep-modules installs dependency-cruiser; migrate-to-shoehorn installs @total-typescript/shoehorn. These are fully documented and intentional, but users should understand that invoking these skills will modify their project dependencies.
Evidence: setup-pre-commit SKILL.md: 'Install as devDependencies: husky lint-staged prettier'; setup-ts-deep-modules SKILL.md: 'Install dependency-cruiser as a devDependency'
LOW wizard skill generates credential-handling shell scripts -5 ▶
The wizard skill generates bash scripts that interact with .env files and GitHub secrets. While designed for legitimate guided credential setup and explicitly does not auto-execute, users should be aware of what scripts are generated before running them.
Evidence: wizard SKILL.md: 'captures what comes back, and writes it into .env files and GitHub Actions secrets'; 'The agent writes the script; it never runs it. You do, on your own machine.'
LOW research skill has documented multi-agent spawning bug -10 ▶
The research skill's own documentation acknowledges issue #530 where the background sub-agent can re-delegate and spawn additional agents, leading to excessive token consumption (~450k tokens reported in one case). This is a resource/cost issue, not a security attack.
Evidence: docs/engineering/research.md: 'This is an open bug, issue #530. The skill tells its caller to spin up a background agent but does not restrict the agent type... one reporter measured a single research task costing roughly 450k tokens across three overlapping runs'
INFO Clean clone with single expected GitHub connection 0 ▶
The git clone operation connected only to GitHub (140.82.121.3:443), consistent with the declared repository URL. No additional outbound connections were observed.
Evidence: Network capture: 'ens3 Out IP 10.0.2.15.47088 > 140.82.121.3.443: tcp'; DNS resolution for github.com
INFO Canary file accesses preceded git clone — attributed to audit infrastructure 0 ▶
Access events for .env, .ssh/id_rsa, .aws/credentials, .npmrc, .docker/config.json, and gcloud credentials were recorded at timestamp 1789093793, approximately 5 seconds before the git clone at 1789093798. This timing is consistent with the audit infrastructure performing its baseline canary enumeration before any skill code ran. All files remain intact.
Evidence: auditd PATH msg=audit(1789093793.267): .env, .ssh/id_rsa, .aws/credentials — clone EXECVE msg=audit(1789093798.788): git clone
INFO Skills use transparent agent delegation pattern 0 ▶
Multiple skills compose by instructing 'Call the Skill tool with X' (e.g., grill-me calls grilling, grill-with-docs calls grilling and domain-modeling). This is the intended skill composition mechanism documented in the plugin architecture, not an attempt to chain unauthorized skills.
Evidence: grill-me SKILL.md: 'Call the Skill tool with grilling.'; grill-with-docs SKILL.md: 'Call the Skill tool twice, for grilling and domain-modeling'
Should I install mattpocock/skills?
Oathe's verdict for mattpocock/skills is SAFE with a trust score of 90/100. Recommendation: Install.