Grep

Bundled with Unleash skills pack. Source: C:\Users\Admin.agents\skills\grep\SKILL.md

grep is IDASQL's entity-search surface. Use it to discover named functions, labels, segments, structs, enums, and members before pivoting into xrefs, decompiler, or type work.

Trigger Intents

Use this skill when user asks to:

• find functions, labels, types, or members by name
• search by prefix/substring like sub_, EH, Zw, CreateFile, or main
• page through search results quickly
• seed table-native xref/decompiler/type workflows from name discovery

Route to:

• xrefs after locating a candidate callee/import/function and needing callers/callees/references
• decompiler after choosing a candidate function to inspect semantically
• types when the hit is a struct/enum/member you need to inspect or edit

Do This First (Quick Start)

-- 1) Start with a structured search while you learn the result shape
SELECT name, kind, addr
FROM grep
WHERE pattern = 'main'
ORDER BY kind, name
LIMIT 20;

-- 2) Narrow immediately when the result set is noisy
SELECT name, ordinal, full_name
FROM grep
WHERE pattern = 'EH%' AND kind = 'struct'
ORDER BY name;

-- 3) Page with ordinary SQL
SELECT name, kind, addr
FROM grep
WHERE pattern = 'sub_%'
ORDER BY kind, name
LIMIT 10 OFFSET 10;

Interpretation guidance:

• grep is a table. Use normal SQL for filtering, sorting, joining, grouping, and paging.
• For downstream parsing, select rows directly instead of wrapping the results in JSON.

Result Shape

grep exposes named IDA entities as rows:

• name
• kind
• addr
• ordinal
• parent_name
• full_name

Common kind values:

• function
• label
• segment
• struct
• union
• enum
• member
• enum_member

Pattern Rules

• Matching is case-insensitive.
• Plain text becomes a contains-match.
• % matches any substring.
• _ matches a single character.
• * is accepted and normalized to %.
• Empty pattern returns no rows.
• This is not regex.
• This is unrelated to byte_search.

Examples:

-- Contains-match
SELECT name, kind
FROM grep
WHERE pattern = 'main'
LIMIT 20;

-- Prefix wildcard
SELECT name, kind, addr
FROM grep
WHERE pattern = 'sub_%'
ORDER BY name
LIMIT 20;

-- Shell-style star is accepted too
SELECT name, kind
FROM grep
WHERE pattern = 'Zw*'
LIMIT 20;

Common Workflows

Find candidate functions by name

SELECT name, addr
FROM grep
WHERE pattern = 'main%' AND kind = 'function'
ORDER BY name;

Resolve imported APIs

SELECT module, name, addr
FROM imports
WHERE name LIKE 'CreateFile%'
ORDER BY module, name;

Find types by convention

SELECT name, kind, ordinal, full_name
FROM grep
WHERE pattern = 'EH%' AND kind IN ('struct', 'enum')
ORDER BY kind, name;

Find members under a parent type

SELECT name, parent_name, ordinal
FROM grep
WHERE pattern = 'flag%' AND kind = 'member'
ORDER BY parent_name, name
LIMIT 30;

Join into richer function metadata

SELECT g.name, f.size, f.prototype
FROM grep g
JOIN funcs f ON f.addr = g.addr
WHERE g.pattern = 'sub_%' AND g.kind = 'function'
ORDER BY f.size DESC
LIMIT 20;

Pivot from discovery into xrefs

SELECT caller_name, printf('0x%X', caller_addr) AS from_addr
FROM callers
WHERE func_addr = (
    SELECT addr
    FROM imports
    WHERE name = 'CreateFileW'
    ORDER BY name
    LIMIT 1
);

Compare With Other Search Surfaces

• Use grep for named entities discovered by IDA.
• Use strings when you need literal string contents.
• Use byte_search when you need raw bytes or opcode patterns.
• Use xrefs after discovery when the real question is "who references this?"

Failure and Recovery

• Too many hits: add kind = ..., tighten the prefix, or switch from plain text to a more specific wildcard pattern.
• No hits for an expected symbol: broaden the pattern, try a contains search, or pivot to imports if the target may only exist as an imported API.
• Need to search for comments, pseudocode text, or string contents: grep is the wrong surface; pivot to strings, decompiler tables, or other domain tables.
• Need bytes/opcodes: use byte_search instead of grep.