Skip to content

Security: hacktheseo/wordpress-seo-skills

Security

SECURITY.md

Security

Reporting a vulnerability

Email security@hacktheseo.com with the skill name, what you did, and what happened. Please do not open a public issue for anything exploitable.

We aim to acknowledge within three working days and to ship a fix or a documented mitigation within fourteen. If you want credit in the changelog, say so.

What these skills do and do not do

An Agent Skill is instructions plus code that runs on your machine, with whatever access your agent has. That is worth taking seriously, including with ours.

These skills do:

  • Read files you point them at: access logs, CSV exports, HTML you saved.
  • Fetch public URLs you name, and treat what comes back as data.
  • Write output files into your working directory.
  • Run Python scripts from their own scripts/ folder, standard library only.
  • Perform DNS lookups, in one place. ai-bot-log-forensics does a reverse then forward DNS resolution on the IP addresses found in your log, because that is the only way to tell a real GPTBot from anything that types GPTBot in a header. It is a DNS query against your own resolver, it carries no payload, and it is the entire point of the verification step. Run the parser with --no-dns to skip it, and you get an unverified count you should not put in a client report.

These skills do not:

  • Send anything anywhere. There is no telemetry, no analytics, no phone home. Apart from the DNS lookups above, the only network traffic is fetching a URL you explicitly asked to audit, and shared/fetch_ranges.py, which you run on purpose and which downloads only the providers' own published IP range files (openai.com, claude.com, google.com, perplexity.ai, apple.com, bing.com). It sends nothing, and it never invents a prefix: a source that fails to answer leaves that provider empty, so its hits are reported as unverifiable.
  • Contain any secret, key, token or internal hostname.
  • Install anything. No pip install, no downloaded binary, no package registry.
  • Write outside your working directory, and never into ~/.claude/, .git/hooks/, shell profiles, or any agent configuration file.
  • Modify your WordPress site. Everything here is read only. The write path lives in the plugin, behind its own permissions, not in a skill.

Prompt injection

Two of these skills read content that someone else controls: a fetched web page, an access log where the user agent and the requested path are attacker-supplied strings.

Every such input is treated as data, never as instructions. This is stated inside the skills themselves so the agent reading them inherits the rule, and the log parser escapes what it emits. If you find a path where fetched or logged content changes the agent's behaviour, that is a vulnerability and we want to hear about it.

Verifying before you install

git clone https://github.com/hacktheseo/wordpress-seo-skills.git
cd wordpress-seo-skills
# every network and shell call, word boundaries so "evals" and "executed" do not match
grep -rnE "\burllib\b|\brequests\.|\bsocket\.|\bsubprocess\b" --include=*.py skills/ shared/
# nothing should match: these are the patterns tools/verify.py itself refuses
grep -rnE "\beval\(|\bexec\(|\bbase64\b|__import__|os\.system\(|shell\s*=\s*True" --include=*.py skills/ shared/
find . -name "*.py" | xargs wc -l                                  # it is all readable Python

Then read SKILL.md for the skills you plan to use. They are under 300 lines each, on purpose.

Windows

Use py -3 in place of python3. If neither works, install Python from python.org and tick "Add python.exe to PATH" during setup.

A general note

In an audit of 3 984 published skills in February 2026, 36,8 % had at least one security flaw, 10,9 % contained a hard-coded secret, and 76 malicious payloads were confirmed by hand. Install skills from sources you can identify, and read them first. That advice includes this repository.

There aren't any published security advisories