agentleFS
HainrixzHainrixz/claude-db/skills/introspect/SKILL.md

introspect

Skill22 starsChanged 5 months agoMITView on GitHub

Tier-1 live read-only introspection. Connects to the running database through a DB MCP server's read-class tools (preferred) or a least-privilege read-only connection string, reads catalogs only (no table data, no writes), and upgrades affected findings to confidence established. Enforces the read-only contract; when no live access is available it marks the dependent checks needs_api — never a silent pass. Invoked by db-orchestrator and the auditors when a check needs runtime truth.

4 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 0 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Else a least-privilege read-only connection string: read from $DATABASE_URL in the environment — never echoed, never written into a finding.

    line 14
  2. Very specific:

    only SELECT / EXPLAIN (no ANALYZE write side effects beyond plan) / catalog reads.

    line 20

“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md
<!-- Copyright (c) 2026 Enrique Rocha (@soyenriquerocha) · tododeia.com · id er-f37b7121 -->

introspect (Tier-1 — live read-only)

Tier 0 reads files. Many checks need runtime truth a file cannot give: real index usage, FK-without-index joins, table/row sizes, RLS state, extensions, engine version, autovacuum/wraparound state. introspect is the only path to that, and it is read-only by contract. See references/data-tiers.md.

Connection path (prefer MCP)

  1. DB MCP server (preferred) — cleaner permission boundary. Use only read-class tools whose names match *query* / *read* / *list* / *describe* / *schema*. A generic write-capable query tool is routed through the same read-only validator below; a PreToolUse hook on mcp__.* backs this up. List available tools first; never call a tool that mutates.
  2. Else a least-privilege read-only connection string read from $DATABASE_URL in the environment — never echoed, never written into a finding. redactSecrets() scrubs any credential that reaches a finding, report, or log.
  3. Neither available → emit status: needs_api on every dependent check with the reason and the minimal grant needed. Never silently pass and never fabricate row counts, sizes, or plans.

Read-only contract (enforced before any query)

  • SET default_transaction_read_only = on;
  • a bounded statement_timeout (e.g. SET statement_timeout = '15s';).
  • only SELECT / EXPLAIN (no ANALYZE write side effects beyond plan) / catalog reads. Reject anything else.

Catalogs read (no row data)

  • Postgres: version(), information_schema, pg_catalog, pg_class, pg_index, pg_indexes, pg_constraint, pg_policies, pg_extension. Tier-2 adds pg_stat_user_indexes / pg_stat_user_tables, pg_stat_statements, age(datfrozenxid).
  • MySQL/MariaDB: information_schema (TABLES, STATISTICS, KEY_COLUMN_USAGE), SHOW ENGINE INNODB STATUS (read).
  • Mongo: db.stats(), $indexStats, $collStats (Tier-2).
  • Cassandra: system_schema, nodetool tablehistograms (Tier-2).

Effect on findings

A check answered with live evidence is upgraded to confidence: established and may therefore cap a score (e.g. confirmed int4/serial PK near exhaustion, TXID wraparound imminent, RLS off on a relied-on table). Without live access those same checks stay directional or become needs_api — they never cap on static signal alone. Every finding still conforms to schema/finding.schema.json; verification.reproduce references $DATABASE_URL, never a literal credential.

Output

Return the catalog facts the calling auditor needs (index inventory, RLS state, version, sizes) plus a clear note of which checks remain needs_api and why. Respond in the user's language (EN/ES).

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.