Nonprofit Impact Orchestra

The name and description (donor-ready proposal packages, RBM/MEAL, logframes, budgets, compliance checks) match the SKILL.md workflow, outputs, and rules. There are no unrelated required binaries, environment variables, or config paths. Minor metadata mismatch: registry metadata lists version 1.0.15 while _meta.json contains version 1.0.11 and a numeric publishedAt; this looks like a housekeeping inconsistency but does not affect capability alignment.

SKILL.md stays on-task: parsing project inputs, building program logic, MEAL/GESI/safeguarding, budgets, and producing structured outputs with explicit evidence rules. It explicitly forbids fabricating baselines, partner commitments, or citations and requires confidence labels and retrievable source metadata. The instructions do not ask the agent to read local files, access unrelated system state, or transmit data to arbitrary external endpoints. Note: as an instruction-only skill, actual retrieval of sources or live verification depends on the host agent's network/search capabilities — the skill expects retrievable sources but doesn't itself implement fetching.

No install spec and no code files — the skill is instruction-only, so it won't write or execute new artifacts on the host. This is the lowest-risk install profile.

The skill declares no required environment variables, no credentials, and no config paths. The requested capabilities are proportional to the stated purpose (drafting and structuring proposals) and do not request unrelated secrets or system access.

Flags show always:false and user-invocable:true (defaults). The skill does not request permanent presence or elevated privileges and does not instruct modifying other skills or system-wide settings.