# Cleared Resume Bullets Five patterns for writing about classified work without violating _Five resume bullet patterns for cleared engineers, each with a fictional bad example and a fictional sanitized good example. Use the closing checklist before any bullet goes public._ Used in: Lesson cleared/06 — https://vetslearn.com/lessons/cleared/06 Source: https://vetslearn.com/resources/templates/career#t-cleared-resume-bullets --- # Cleared Resume Bullets **Used in:** [cleared/06 — Talking about your work without violating](/lessons/cleared/06) ## How to use You worked on classified programs. You cannot say what, where, or for whom. You still need a resume that gets you the next job. Five patterns below. Each shows the structure, a bad example, and a good example. The bad examples are fictional. The good examples are fictional. Do not copy them. Use them as a shape and write your own. When in doubt, run your final draft past your FSO before you post. --- ## Pattern 1 — Problem type + customer type + scale + impact **The pattern.** Describe the *kind* of problem, the *type* of customer, the *scale* of the work, and the *outcome*. Never the specific customer, program name, or location. **Bad example (fictional):** > Built RAG pipeline for NSA Project HARVEST that searched 4M classified documents across SIPR and JWICS for the SIGINT directorate. Why it is bad: names the agency, names the program, names the network, names the directorate. Four violations in one sentence. **Good example (fictional):** > Built retrieval system over millions of documents for an intelligence community customer on a high-side network. Cut analyst search time by 60%. Why it works: "intelligence community customer" describes the type. "High-side network" describes the environment without naming it. Scale is real, impact is real. --- ## Pattern 2 — Technology + role + team size + outcome **The pattern.** Lead with the technology stack and what you did with it. The customer is generic. The team size shows scope. **Bad example (fictional):** > Lead engineer on $40M CIA contract building Llama 3 70B inference cluster on AWS GovCloud TS region for OSD/AT&L analysts. Why it is bad: names the agency, names the dollar value (often FOUO), names the specific cloud region, names the end customer office. **Good example (fictional):** > Led 6-engineer team deploying open-weights LLM inference on a classified cloud environment. Served thousands of analysts across a federal customer. Why it works: technology is generic enough. Team size is real. "Classified cloud environment" is fine. "Federal customer" is fine. --- ## Pattern 3 — Compliance framework + accomplishment **The pattern.** When your work involved a public framework (FedRAMP, NIST, STIGs, DISA), you can name the framework. You cannot name the specific system that went through it. **Bad example (fictional):** > Took ODNI's IC GPT system through FedRAMP High ATO in 9 months. Why it is bad: names the agency and the system. **Good example (fictional):** > Led FedRAMP High ATO for a generative AI platform. Authored SSP, ran control assessments, achieved authorization in under 12 months. Why it works: framework is public, accomplishment is concrete, system identity is hidden. --- ## Pattern 4 — Skill demonstrated + measurable outcome **The pattern.** Lead with a skill the next employer cares about. Show the outcome. Skip the program details entirely. **Bad example (fictional):** > Wrote prompt injection detection for NGA's geospatial AI assistant after a red team finding in Q3 2024. Why it is bad: names the agency, names the system, dates the incident. **Good example (fictional):** > Designed prompt injection defenses for a production LLM application. Reduced successful injection rate from 14% to under 1% on internal red team tests. Why it works: skill is hireable, numbers are real, no program identity. --- ## Pattern 5 — Clearance + access without naming the program **The pattern.** Your clearance level is a credential. You can list it. You cannot list the specific compartments or programs unless the program is publicly acknowledged. **Bad example (fictional):** > TS/SCI with CI poly. Read on to TALENT KEYHOLE and BYEMAN. Current on Project STELLAR access. Why it is bad: names compartments, names programs. **Good example (fictional):** > TS/SCI with CI polygraph. Current SCI accesses with multiple intelligence community customers. Why it works: clearance level is fine to state. "Current SCI accesses" signals you are read on somewhere without saying where. --- ## Closing checklist Before you publish or send any bullet, answer yes to all six: 1. Have I removed every program name, code word, and project nickname? 2. Have I removed every specific agency, directorate, or office name? 3. Have I removed every specific location (building, base, city when paired with mission)? 4. Have I removed every specific dollar value that is not publicly available? 5. Have I checked that combining this bullet with my other bullets does not create a mosaic that identifies a program? 6. If this bullet went on LinkedIn tomorrow, would my FSO have a problem with it? If any answer is no, rewrite. If you are not sure, ask your FSO. "I asked first" is a much better story than "I posted and got called in."