phrazzld/oracle
skills/oracle/SKILL.md
Browser-mode-only Oracle consults: bundle a prompt plus selected files and ask a signed-in ChatGPT GPT-5.5 Pro browser session for a second opinion. Use when stuck, debugging hard bugs, reviewing an architecture plan, or cross-checking a substantive diff with large file context. Never use Oracle API mode from Harness Kit. Trigger: /oracle, /consult.
$ install --global
skillsauthnpx skillsauth add phrazzld/agent-skills oracleInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Jul 2, 2026, 6:15 AM285.0s2 files scanned
SKILL.md
- name:
- oracle
- description:
- |
- Browser-mode-only Oracle consults:
- bundle a prompt plus selected files and ask
- API mode from Harness Kit. Trigger:
- /oracle, /consult.
- argument-hint:
- [prompt + file globs]
/oracle
Use Oracle only as a browser-backed consult surface. It bundles the prompt and
selected files, opens a signed-in ChatGPT browser session, and stores the
answer under ~/.oracle/sessions. The lead agent still owns synthesis,
verification, and final judgment.
Hard Boundary
- Always pass
--engine browser. - Prefer
--model gpt-5.5-prounless the operator names another browser model. - Never run
--engine api,--provider openai,--no-azure,--models, or API preflight from this skill. - If an Oracle command would need an API key or per-token billing, do not use Oracle for that work from Harness Kit; use the normal roster/research paths instead.
Workflow
-
Pick the tightest file set that contains the truth. Exclude tests, snapshots, generated output, and fixtures unless they are the oracle.
-
Preview first:
npx -y @steipete/oracle --engine browser --model gpt-5.5-pro \ --dry-run summary --files-report \ -p "<task>" --file "src/**" --file "!**/*.test.*" -
Run only after the preview is sensible:
npx -y @steipete/oracle --engine browser --model gpt-5.5-pro \ --slug "<3-5-words>" \ -p "<task>" --file "src/**" --file "!**/*.test.*" -
If the run detaches or times out, do not re-run. Inspect the stored session:
npx -y @steipete/oracle status --hours 72 npx -y @steipete/oracle session <id> --render
Good Prompts
Oracle starts cold. Include the repo goal, relevant commands, exact failure, constraints, and desired output shape. For review, give the artifact and oracle, not the author's reasoning trail.
Good output requests:
Return blockers only, with file/line evidence.Challenge this plan; list missing proof, wrong assumptions, and safer alternatives.Find the likely root cause and the smallest verification loop.
Safety
- Do not attach secrets,
.env, key files, tokens, browser profiles, or private customer data. - Use
--dry-run summary --files-reportbefore every real run. - Treat Oracle output as advisory evidence. Verify with the repo's live oracle, tests, gate, browser route, or command before claiming done.
- Browser mode is cheaper than API billing but can be slow and subject to ChatGPT subscription limits.
Related Skills
phrazzld/compound
testing
VerifiedTrustedCommunity
Capture one compounding repo-technical learning while a solved problem is still fresh. Use when: after a bug fix, diagnosis, delivery, review, or incident reveals a reusable pattern worth adding to `docs/solutions/`. Trigger: /compound, /capture-learning, /learning.
13SKILL.mdUpdated Jul 5, 2026
phrazzld/factory-apps
testing
VerifiedTrustedCommunity
Route Misty Step factory application capabilities. Use when choosing, auditing, integrating, or operating Canary, Powder, Landmark, Aesthetic, or Bitterblossom: production observability, incidents, health checks, error logging, backlog/work-card state, release intelligence, UI/UX system adoption, or supervised/unsupervised agent dispatch. Trigger: /factory-apps, /factory-stack.
13SKILL.mdUpdated Jul 4, 2026
phrazzld/skill-eval
testing
VerifiedTrustedCommunity
Prove a skill beats no-skill with a falsifiable A/B eval, or retire it. Design, generate, run, and maintain a skill-specific eval: name the one claim the skill must earn, run it skill-on vs raw same-model, grade blind with objective checks first, return a keep/adapt/cut verdict. Use when: "eval this skill", "does this skill help", "prove the skill beats no skill", "write an eval for", "benchmark a skill", "is this skill worth it", "skill A/B", "skill regression test", "generate skill evals". Trigger: /skill-eval, /eval-skill, /prove-skill.
13SKILL.mdUpdated Jul 2, 2026
phrazzld/skills/harness-engineering/templates/repo-local-skill
tools
VerifiedTrustedCommunity
> Template. Copy to `<target-repo>/.agents/skills/<repo>-<domain>/SKILL.md` > and fill every bracketed placeholder from the live target repo. Delete this > line and every other `> ` guidance line before committing. See > `../../references/repo-local-skill-generation.md` for the full process. --- name: <repo>-<domain> description: | [One paragraph: what this skill verifies/runs/operates for <repo>, stated in terms of the repo's real shape (service/CLI/library/etc.), not generic process. En
13SKILL.mdUpdated Jul 2, 2026