Storage

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

netnode_kv

Persistent key-value store backed by IDA netnodes. Data is saved inside the IDB automatically. Supports full CRUD and O(1) key lookup via WHERE key = '...'.

-- Store a value
INSERT OR REPLACE INTO netnode_kv(key, value) VALUES('author', 'alice');

-- Read by key (O(1) lookup)
SELECT value FROM netnode_kv WHERE key = 'author';

-- List all entries
SELECT * FROM netnode_kv;

-- Update a value
UPDATE netnode_kv SET value = '2.0' WHERE key = 'version';

-- Delete an entry
DELETE FROM netnode_kv WHERE key = 'author';

Use Cases

• Session state: Track analysis progress across sessions (e.g., which functions have been annotated)
• Metadata: Store custom metadata like analyst name, analysis date, notes
• Bookkeeping: Track which functions have been re-sourced, annotated, or reviewed
• Configuration: Store per-database analysis settings

-- Track analysis progress
INSERT OR REPLACE INTO netnode_kv(key, value) VALUES('annotated_funcs', '["main","init_config"]');

-- Update progress
UPDATE netnode_kv SET value = '["main","init_config","process_input"]'
WHERE key = 'annotated_funcs';

-- Read progress in a new session
SELECT value FROM netnode_kv WHERE key = 'annotated_funcs';

Performance Rules

Key rules:

• Exact key lookup (WHERE key = '...') is O(1) — this is the preferred access pattern.
• Prefix scans (LIKE 'prefix%') iterate all entries but are fast for typical netnode sizes.
• netnode_kv is stored inside the IDB file — it persists automatically with save_database().

Advanced Storage Patterns

JSON-based progress tracking

Store structured analysis state as JSON for richer querying:

-- Store progress with structured metadata
INSERT OR REPLACE INTO netnode_kv(key, value)
VALUES('progress:overview', json_object(
    'total_funcs', (SELECT COUNT(*) FROM funcs),
    'named_funcs', (SELECT COUNT(*) FROM funcs WHERE name NOT LIKE 'sub_%'),
    'timestamp', datetime('now')
));

-- Read and parse progress
SELECT json_extract(value, '$.total_funcs') AS total,
       json_extract(value, '$.named_funcs') AS named,
       json_extract(value, '$.timestamp') AS ts
FROM netnode_kv WHERE key = 'progress:overview';

Per-function annotation status tracking

Track which functions have been annotated and what was done:

-- Mark a function as annotated
INSERT OR REPLACE INTO netnode_kv(key, value)
VALUES('re_source:' || printf('0x%X', 0x401000),
       json_object('status', 'done', 'summary', 'DriverEntry init',
                    'analyst', 'alice', 'date', date('now')));

-- Find unannotated functions by joining with funcs
SELECT f.name, printf('0x%X', f.addr) AS addr
FROM funcs f
WHERE f.name NOT LIKE 'sub_%'
  AND NOT EXISTS (
    SELECT 1 FROM netnode_kv
    WHERE key = 're_source:' || printf('0x%X', f.addr)
  )
ORDER BY f.size DESC
LIMIT 20;

Naming conventions for keys

Use a namespace:entity:id format for organized storage:

re_source:0x401000          → per-function annotation status
config:string_minlen        → analysis configuration
snapshot:2024-01-15          → point-in-time analysis snapshot
tag:crypto:0x401000         → function tags/categories

-- List all keys in a namespace
SELECT key, value FROM netnode_kv WHERE key LIKE 'tag:crypto:%';

-- Count entries per namespace
SELECT SUBSTR(key, 1, INSTR(key, ':') - 1) AS namespace,
       COUNT(*) AS entries
FROM netnode_kv
WHERE key LIKE '%:%'
GROUP BY namespace
ORDER BY entries DESC;

See Also

• annotations — for analyst breadcrumbs that should live in the address space (comments, bookmarks). Use netnode_kv for cross-address session state.