Skill Claude Code
Model adversarial users, attackers, malformed inputs, prompt injection, abuse cases, and metric gaming.
Skill Claude Code
Model adversarial users, attackers, malformed inputs, prompt injection, abuse cases, and metric gaming.
Skill Claude Code
Check whether APIs, CLIs, and UIs make the right actions obvious and wrong actions hard.
Skill Claude Code
Apply Talebian reasoning to software. expose fragility, tail risk, false confidence, harmful intervention, and opportunities for via negativa, optionality, barbell design, and antifragile feedback.
Skill Claude Code
Treat deliberate deletion as a healthy biological function. Apply apoptosis vs necrosis, autophagy, senescence, and the principle that not-dying is pathological to deprecation, sunsetting, and pruning.
Skill Claude Code
Surface hidden assumptions in code, specs, designs, and debugging plans; classify which are guaranteed, validated, trusted, inherited, or unsafe.
Skill Claude Code
Review notifications, prompts, logs, errors, and UI signals for salience, interruption cost, and prioritization.
Skill Claude Code
Update beliefs about bugs, hypotheses, and triage decisions using priors, likelihoods, and posteriors rather than gut anchoring on the latest piece of evidence.
Skill Claude Code
Check reasoning for anchoring, confirmation bias, availability bias, sunk-cost fallacy, and premature closure.
Skill Claude Code
Behavioral finance applied to tech adoption — recognize hype cycles, herding, FOMO, and Minsky moments before deciding to chase the next framework, paradigm, or migration.
Skill Claude Code
Decide where to spend engineering effort across competing demands using ROI, hurdle rates, opportunity cost, and zero-based budgeting — and ignore sunk cost.
Skill Claude Code
Reconstruct timelines and causal chains from logs, stack traces, commits, artifacts, tests, and changed files before deciding what happened.
Skill Claude Code
Improve readability, conceptual flow, naming, API story, and maintainability of complex code.
Skill Claude Code
Evaluate code, APIs, docs, and workflows for working-memory burden, chunking, naming, and mental model fit.
Skill Claude Code
Apply engineering bring-up and tear-down protocols (FAT, SAT, punchlist, LOTO, mothballing) to service launches and sunsets so systems start clean and end cleanly.
Skill Claude Code
Grice's maxims and speech act theory applied to engineering text — error messages, commit messages, log lines, prompts, and PR descriptions. Distinguishes literal meaning from intended effect.
Skill Claude Code
Find the current bottleneck, avoid optimizing non-constraints, and reason about throughput and queues.
Skill Claude Code
Apply epidemiological reasoning — R0, generation time, susceptibility, contact tracing, herd immunity — to vulnerability propagation, dependency CVEs, credential leaks, bug spread, and incident blast radius.
Skill Claude Code
PID controllers and classical control theory applied to autoscalers, congestion control, rate limiters, and any feedback-driven regulator. Use when a system must drive a measured variable to a setpoint without oscillating, overshooting, or stalling.
Skill Claude Code
Use nucleation, supersaturation, and crystal growth to reason about standards adoption, pattern propagation, and why some ideas suddenly "take" while others remain dissolved.
Skill Claude Code
Use differential diagnosis to triage possible causes by likelihood, severity, and test cost.
Skill Claude Code
Track units of measure across code, APIs, and data so unit-mismatch operations are caught at the type, spec, or review level rather than at runtime.
Skill Claude Code
Analyze how knowledge is shared across code, docs, tools, tests, teams, and rituals.
Skill Claude Code
Look for emergent behavior where simple local rules produce surprising global effects.
Skill Claude Code
Act as a senior engineering coach. Review an engineer's recent work across PRs, code reviews, issues, and commits; produce an evidence-cited SWOT and a short list of prioritized, behavioural actions — with room for the engineer to add context and challenge findings.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: