Public leaderboard

Public assessment

dmang-dev/mcp-mgba (mcp-mgba)

mcp-mgba · v0.3.0 · scanned

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.

01Safety

0.0 / 30

0.0 out of 30
02Legibility

28.4 / 30

28.4 out of 30
03Economics

15.6 / 20

15.6 out of 20
04Discoverability

8.3 / 20

8.3 out of 20

Highest-impact fix

Estimated gain +30 points

Add 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.

10 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 read a numeric field at an address; a task like 'read the score value' doesn't specify bit width, so the agent could pick either without checking whether the game field is actually 16- or 32-bit.
mgba_read16 mgba_write16 low Distinct read/write verbs are usually clear, but a vague task like 'check/set HP at this address' could momentarily blur which operation is intended before the agent notices the verb.
mgba_read32 mgba_write32 low Same read/write distinction as read16/write16; verbs disambiguate but a terse instruction referencing only the address and width could cause hesitation.
mgba_write16 mgba_write32 medium A task like 'set the player's coordinates to X' without specifying width leaves ambiguity between 16-bit and 32-bit write, especially since GBA often uses 32-bit pointers/values interchangeably with 16-bit fields in different games.
mgba_read8 mgba_read16 low Both read small values from an address; if the task doesn't mention 'byte' vs a specific bit width, the agent might default to the wrong granularity for a status flag vs counter.
mgba_read8 mgba_write8 low Clear read/write verbs typically disambiguate, but a combined instruction like 'inspect and set this flag' could cause momentary tool confusion.
mgba_write8 mgba_write_range medium A task like 'write these bytes to memory' without specifying count could lead the agent to choose write8 (single) vs write_range (multiple) incorrectly if it misjudges how many bytes are involved.
mgba_pause mgba_unpause low Opposite actions with clear naming, but a task phrased ambiguously like 'toggle emulation state' without saying which direction could cause a wrong pick.
mgba_save_state mgba_load_state low Clear save/load semantics, but a task like 'restore/checkpoint the game state' phrased ambiguously (e.g., 'get me back to state in slot 3') could momentarily be misrouted if intent verb is unclear.
mgba_read_range mgba_write_range low Distinct read/write purpose, but a workflow instruction like 'capture and restore this memory region' bundles both, risking the agent picking only one or confusing which is needed for 'restore' vs 'inspect'.

Compare the field

One score is useful.
The evidence makes it actionable.

Back to the leaderboard