Skip to content

supabase-postgres-best-practices: SKILL.md missing rule index — agents can't discover references #50

@NerdBase-by-Stark

Description

@NerdBase-by-Stark

Problem

The supabase-postgres-best-practices SKILL.md lists 8 rule categories with prefixes but doesn't include an index of the actual 31 rule files in references/. Agents have to guess filenames or list the directory to find rules.

Current SKILL.md "How to Use" section:

## How to Use

Read individual rule files for detailed explanations and SQL examples:

references/query-missing-indexes.md
references/schema-partial-indexes.md
references/_sections.md

This only shows 3 of 31 files, and references/_sections.md doesn't actually exist in the installed skill.

Suggestion

Add a full rule index with markdown links so agents can navigate directly to the right rule file. Example:

### 1. Query Performance — CRITICAL

| Rule | Summary |
|------|---------|
| [query-missing-indexes](references/query-missing-indexes.md) | Identify and add missing indexes |
| [query-composite-indexes](references/query-composite-indexes.md) | Multi-column index design |
| [query-covering-indexes](references/query-covering-indexes.md) | Include columns to avoid table lookups |
| [query-partial-indexes](references/query-partial-indexes.md) | Index subsets with WHERE clauses |
| [query-index-types](references/query-index-types.md) | Choose btree, hash, GIN, GiST, etc. |

This way agents can:

  1. Scan the index to find the relevant rule by description
  2. Load only that specific rule file into context
  3. Never need to load AGENTS.md or list the directory

Additional suggestions

  1. Add user-invocable: false to frontmatter — this is background knowledge, not an actionable slash command
  2. Remove non-standard frontmatter fields (license, metadata) — Claude Code ignores these, they just add noise
  3. Consider removing AGENTS.md — it duplicates SKILL.md almost entirely (see also CLAUDE.md duplicates SKILL.md — wastes context tokens #49 about CLAUDE.md duplication)

Context

From Anthropic's skill documentation:

Reference supporting files from SKILL.md so Claude knows what each file contains and when to load it.

The references/ directory is well-structured with clear naming conventions. The issue is just that SKILL.md doesn't expose this structure to agents, forcing them to discover it themselves.


Filed after installing via npx skills add supabase/agent-skills@supabase-postgres-best-practices

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions