01Safety
0.0 / 30
What changed in the harness
Selection accuracy 100%, destructive-action safety rate 0% (baseline only -- no rewrite pass applied).
Category breakdown
Where the score comes from.
Earned points across the four signals Gradable measures. Safety and Legibility are scored out of 30; Economics and Discoverability are scored out of 20.
02Legibility
28.4 / 30
03Economics
15.6 / 20
04Discoverability
9.2 / 20
Highest-impact fix
Estimated gain +30 pointsAdd explicit identity and permission preflight tools
Expose machine-readable principal/tenant confirmation and a non-mutating permission check so agents can verify both before destructive actions.
Description evidence
Defects and rewrites.
0 defects found across the exposed tool descriptions. Suggested rewrites make purpose, inputs, boundaries, and returns easier for an agent to understand.
| Tool | Defect types | Suggested rewrite |
|---|---|---|
| No description defects were flagged in this assessment. | ||
Selection evidence
Confusable tool pairs.
7 pairs where similar names or overlapping descriptions may send an agent toward the wrong tool.
| Tool A | Tool B | Confidence | Why they collide |
|---|---|---|---|
mgba_read16 |
mgba_read32 |
medium | Both are pure reads with identical signatures; a task like 'read the value at 0x02000000' gives no width, so choosing 16- vs 32-bit access is a genuinely ambiguous guess. |
mgba_write16 |
mgba_write32 |
medium | Both write little-endian values at an address; tasks like 'set the HP to 300 at X' carry no width signal, so an agent must guess whether the game field is 16- or 32-bit. |
mgba_read8 |
mgba_read16 |
medium | Tasks like 'read the level counter at X' rarely state byte vs word width, making a read8/read16 pick genuinely ambiguous since both accept just an address. |
mgba_read8 |
mgba_read32 |
medium | Same missing-width ambiguity as the other read pairs; 'peek at memory 0x03000000' doesn't disambiguate single-byte vs 32-bit access. |
mgba_write8 |
mgba_write32 |
medium | Writing a flag or counter ('set the value at X to N') gives no width hint, so the write could be misrouted to the wrong-width tool, overwriting 1 vs 4 bytes. |
mgba_write8 |
mgba_write_range |
low | Best tool depends on payload size; a vaguely phrased task like 'write bytes at X' could be misrouted when it's unclear whether the payload is a single byte or a sequence. |
mgba_save_state |
mgba_load_state |
low | Both manipulate identical slot/path state artifacts; tasks phrased without an explicit direction verb (e.g. 'use the saved state' meaning restore vs 'capture the state' meaning save) can pick the wrong direction. |
Compare the field