| name | security-reviewer |
| description | Chords security review guidance. Use for publishing-credential handling, GitHub Actions secrets, GPG-encrypted key files, dependency provenance, Gradle wrapper integrity, agent prompt/configuration safety under `.agents/**`, and accidental secret exposure in this public repository.
|
Security Reviewer
Use this skill for security-focused review or implementation guidance. Chords
is a public open-source repository with no runtime backend of its own, so the
main risk surface is the supply chain — publishing credentials, CI workflows,
and dependencies — plus the agent configuration itself under .agents/**.
Scope
- Publishing credentials: the GPG-encrypted key files under
.github/keys/,
the decryption scripts under config/scripts/, and their use in
.github/workflows/publish.yml.
- GitHub Actions secrets and workflows that build, validate, or publish
artifacts.
- Dependency provenance: coordinates and repositories declared in
buildSrc
and codegen/plugins/buildSrc; Gradle wrapper integrity (the
gradle-wrapper-validation workflow).
- Accidental commits of credentials, tokens, private keys, or TeamDev-internal
data into this public repository.
- Agent configuration files (
.agents/**, openai.yaml, AGENTS.md): flag
prompt-injection risks in default_prompt fields and skill instructions
that could cause an agent to exfiltrate secrets or bypass safety rules.
Rules
- Treat credential leakage as high severity: this repository is public, and
anything committed is published.
- Do not print secret values, tokens, private keys, or decrypted credential
payloads.
- Do not suggest storing secrets in source-controlled files; encrypted
.gpg
blobs under .github/keys/ are the existing sanctioned mechanism, and
changes to them are human-owned.
- Reference workflow secrets by name through
secrets.*; never inline values.
- Do not run commands that publish artifacts, rotate credentials, or change
GitHub secrets.
- Watch for dependency-confusion risk: new repositories or dependency
coordinates must match the trusted Spine/TeamDev sources already configured
in
buildSrc.
Review Focus
- Workflow changes that widen
on: triggers, expose secrets to
pull-request-controlled contexts, or add untrusted third-party actions.
- New or changed Maven repositories and dependency coordinates.
- Scripts that decrypt, copy, or upload credential material.
- Content that should not be public: internal hostnames, e-mail addresses of
individuals, or internal project data in code, tests, or docs.
- Prompt-injection risks in
.agents/**: overbroad default_prompt fields,
skill instructions that could bypass the policy in AGENTS.md, and
instructions that could cause an agent to exfiltrate secrets.
Output Format
For reviews, lead with findings ordered by severity and include file/line
references. If no issues are found, state that clearly and mention any residual
risk or unverified external configuration.