--- name: policy-grounding description: Answers only from the shop's own documents (its policies, FAQ and product catalog), quotes the passage each claim relies on, and says plainly when the documents do not cover the question. Use whenever a reply depends on what the shop allows or on a fact about a product. ---
policy-grounding
Answers only from the shop's own documents (its policies, FAQ and product catalog), quotes the passage each claim relies on, and says plainly when the documents do not cover the question. Use whenever a reply depends on what the shop allows or on a fact about a product.
- Instant access
- Portable SKILL.md
- Free updates
One file, ready to install
The body stays locked until purchase — here is precisely what unlocks.
Tool permissions not declared
No allowed-tools list, so whatever your runtime permits by default applies.
What you install
A skill directory you can read the head of before you trust it.
- The SKILL.md file itself1 file · 2.3 KB
- Its frontmatter, exactly as publishedReadable in full above before you buy — only the body is held back.
- Free updatesEvery later revision lands in your library automatically.
- SKILL.md 2.3 KB · 530 tk
What it costs to keep loaded
A skill loads in stages — the index entry is always resident, the body only when the skill fires.
Counted with a general-purpose tokenizer, so a particular runtime will differ by a few percent. The skill makes no model call of its own, so there are no output tokens to pay for beyond the reply you were already getting.
Specifics and common questions
- Version
- v1
- Licence
- Not stated
- Allowed tools
- Not declared
- Ships as
- 1 file · 2.3 KB
- Runs in
- Any SKILL.md runtime
- Spec
- agentskills.io v1
- Context cost
- 530 tokens59 always resident · the rest only when the skill fires.
- Published
- Oct 18, 2026
- Reviews
- No reviews yet
Experimental bio with extra words, and extra cofe.
-
What does policy-grounding do?
Answers only from the shop's own documents (its policies, FAQ and product catalog), quotes the passage each claim relies on, and says plainly when the documents do not cover the question. Use whenever a reply depends on what the shop allows or on a fact about a product.
-
Who is policy-grounding best for?
policy-grounding is built for research & analysis use cases and is particularly useful for documents, grounding, and policy.
-
How much does policy-grounding cost?
policy-grounding is free to claim on Sigrix.
-
How do I use policy-grounding?
policy-grounding is delivered as a portable SKILL.md module you own on Sigrix. Drop it into Claude Code, claude.ai, or the Claude API and tailor the instructions to your own workflow.
Report this skill
Tell us what's wrong. A moderator reviews every report.
Thanks -- we got it. A moderator will take a look.
More skills
support-reply
Drafts a reply to a customer email in the shop's voice: acknowledge, answer, give the next step, sign off. Also decides when a person should answer instead, and what the customer is sent while they do. Use for every reply you draft to a customer.
fair-hiring-workflow-small-teams
A practical, repeatable hiring workflow for founders and small teams without dedicated HR. Move from role scoping to a clear job description, structured interviews, consistent evidence-based scorecards, and a documented debrief. Includes prompts, templates, fairness checks, and guidance for keeping human judgment in control—without ranking candidates automatically or inferring protected traits.
sigrix-mcp-setup-doctor
Help Sigrix sellers configure the Sigrix MCP in VS Code, Cursor, Claude Code, or Claude Desktop, verify connectivity, and diagnose authentication and startup failures. Use when the server is missing, seller tools return 401, or setup behaves differently across clients; protect tokens and distinguish public connectivity from authenticated seller access.
sigrix-listing-workshop
Turn a rough idea, existing prompt, persona, or skill into a clear Sigrix listing. Use when planning, writing, refining, or validating marketplace listing content; preserve the author's intent, identify missing decisions, and avoid unsupported claims.