skills/writing-research/grounding-with-google-genai/SKILL.md
Run one-shot Google Gemini research through the official Google GenAI Interactions API with Google Search, Google Maps, or URL Context grounding and inspect citation evidence. Use when Codex needs current web grounding, place or local recommendations, answers about specific public HTTP(S) URLs, or an explicit Gemini/Google GenAI grounded response.
npx skillsauth add narumiruna/agent-skills grounding-with-google-genaiInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Use the bundled scripts/google_genai_grounding.py by absolute path. Default to gemini-3.5-flash; pass --model only when the user requests another compatible model. The script sends stateless requests with store=false and emits JSON containing the answer, citations, bounded tool evidence, and usage.
search for current or open-web facts that benefit from Google Search citations.maps for places, local recommendations, addresses, reviews, or proximity. Supply both coordinates when the user's relevant location is known; omit both otherwise. Google Maps grounding currently requires English prompts and responses, so query in English and translate the cited answer afterward when needed.url to read or compare one to 20 specific public http:// or https:// pages. Prefer direct content URLs. Do not use it for localhost, private networks, login-only or paywalled pages, Google Workspace files, YouTube, or audio/video URLs.Resolve SKILL_DIR to this skill directory and require GEMINI_API_KEY in the process environment or a discoverable .env file. The script loads .env with python-dotenv without overriding an existing environment value. Never print, log, place in arguments, or write the key to repository files.
Run Google Search grounding:
uv run --script "$SKILL_DIR/scripts/google_genai_grounding.py" search \
'What changed in the current release? Cite primary sources.'
Run Google Maps grounding, adding coordinates only when relevant:
uv run --script "$SKILL_DIR/scripts/google_genai_grounding.py" maps \
'Find a well-reviewed coffee shop within a 15-minute walk.' \
--latitude 25.033 --longitude 121.5654
Run URL Context with one --url per source:
uv run --script "$SKILL_DIR/scripts/google_genai_grounding.py" url \
'Compare the documented behavior and identify material differences.' \
--url 'https://example.com/one' \
--url 'https://example.com/two'
A request to use this skill authorizes one bounded grounding call. Grounded calls can consume quota or incur charges; retry once only for a clear transient failure or a corrected deterministic input, and do not fan out into extra calls without need.
Inspect citations, tool_steps, and usage before relying on text:
url_citation or place_citation evidence for grounded claims.status: success. Treat error or unsafe as failed retrieval even if the model produced plausible text.interaction_id is expected for stateless requests.Stop after the grounded answer and its material caveats are complete. Do not broaden a specific URL request into general web research unless the user asks for both.
development
Score or compare one or more agent skills across trigger clarity, workflow actionability, safety boundaries, verification rigor, incremental knowledge value, and leanness. Use only when the user explicitly asks for ratings, numerical quality scores, rubric-based scorecards, or scored comparisons; use creating-agent-skills for unscored reviews or revisions.
development
Assess or improve an existing codebase's architecture when the user asks about module boundaries, coupling, scattered ownership, testability, change locality, deep modules, seams, or behavior-preserving structural refactoring. Use for cross-module design rather than ordinary diff review or a confirmed edge-case bug fix.
development
Perform read-only security audits, vulnerability assessments, or threat-focused reviews of diffs, pull requests, code paths, or explicitly scoped repositories when security is the primary objective or acceptance criterion. Use reviewing-code for ordinary review with baseline security coverage and hardening-code-paths for fixing confirmed findings.
development
Run iterative multi-reviewer panels over a code diff, verify their findings, apply explicitly authorized fixes, and re-review the updated change until it passes or reaches a stopping condition. Use when the user asks for a panel loop, multi-model code-review consensus, or a review-fix-re-review cycle.