Ben Huffer
Community@flyfission · Greenville SC
Founder of FlyFission Consulting Group. Building nuclear-grade AI, software, and engineering workflows for high-consequence projects.
Agent Skills by Ben Huffer
Showing 26 vetted skills indexed across 1 GitHub repositories.
closing-stale-packets
Close stale change packets flagged by ng status with a documented rationale.
declaring-intent
Declare intent before irreversible actions with abort criteria and rollback.
reviewing-code-quality
Review code diffs for standards drift and maintainability risks.
creating-change-records
Automate creation and revision of Nuclear-grade change records with traceability.
checking-legal-and-safety-wording
Rewrite overreaching legal, safety, and compliance claims in public documentation.
rating-change-risk
Determine Nuclear-grade change mode from stakes, reversibility, exposure, unknowns, and proofs.
responding-to-incidents
Declare incidents with commanders, timelines, and owned corrective actions.
vetting-outside-code-and-models
Document external dependency trust decisions with observed evidence and release impact.
checking-release-readiness
Record release decisions with evidence, baseline, risk, rollback, and monitoring.
learning-from-experience
Transform incidents and near misses into no-blame OPEX records with control updates.
handing-off-work
Record unfinished work handoffs into a turnover.md artifact.
organizing-project-folders
Generate a folder hierarchy from a work breakdown with naming and depth rules.
using-nuclear-grade
Automate scoping and routing decisions for AI-assisted software changes.
checking-source-claims
Verify cited sources are public and classify their influence, evidence, or authority.
staying-on-mission
Compare current actions against a written mission anchor and record a re-anchor, escalate, or stop decision.
recording-what-an-agent-did
Record agent-run tool calls, edits, inputs, outputs, tokens, and approvals as auditable trace evidence.
questioning-attitude
Restate a change as a decision question and document assumptions and evidence gaps.
choosing-what-to-control
Identify code, prompts, models, and tools requiring approved states.
stress-testing-agent-changes
Simulate attacks on agent changes and record risk results.
proving-claims
Map claims to evidence rows with statuses and release posture.
deciding-who-decides
Determine edge agent or human gate authority using reversibility, evidence, and consequence.
tracking-deficiencies
Record known deficiencies with first-seen date, owner, and disposition.
recording-a-known-good-version
Record known-good baselines with accepted versions, evidence, and re-record triggers.
checking-what-a-change-affects
Generate a change-impact.md with actions, blockers, and owners for ripple effects.