Drill 1: Compromised Credential / Delegated Token

1. 01

   Concepts & system design

   60 min· read · map the mechanismtimer

   R22 — OAuth revocation

   R23 — Token exchange

   R11 — Model pinning

   R37 — SLSA/Sigstore verification

   Read the assigned sections; map the mechanism, threat model and trade-offs.

2. 02

   Project implementation

   120 min· build · integrate · debugtimer

   Simulate a leaked API key or compromised D7 agent OAuth token in your local lab. Execute immediate token revocation, secret rotation, active-session termination, and blast-radius reconstruction using the D11 Cryptographic Trajectory Ledger.

   Simulate a compromised or silently altered model checkpoint/provider update (D12). Detect hash/behavior drift, block rollout, execute automated rollback to the pinned known-good model version, and rerun the D8 evaluation suite.

   Build, integrate, debug and improve the increment in the active project.

3. 03

   Security lab

   60 min· attack your own systemtimer

   No separate task text for this block on day 95 — the sourced work sits in the block above. Keep the time box.

   Test the abuse case, failure path, authorization bypass or resource limit.

4. 04

   Verification & evidence

   30 min· test · record evidencetimer

   Post-incident drill reports and timeline logs for Drills 1 and 2.

   Add or run regression tests; record the 9-point evidence record.

5. 05

   Deliberate practice & Socratic review

   30 min· explain aloud · update gapstimer

   Answer the 8 Socratic questions without notes: What? Why? How? What if? Why not? How can I prove it? Where can it break? What have I learned?

   Re-explain the hardest concept aloud, inspect one edge case, update the gap log.

6. total5 h

• R11Hugging Face Documentation↗ huggingface.co

  Model cards, tokenizer/config files, embedding usage, and repository commit-hash revision pinning.

• R22OAuth 2.0 Security BCP (RFC 9700)↗ rfc-editor.org

  RFC 6749 core flows + RFC 9700 security best practices: narrow scopes, audience restriction, token leakage, and replay.

• R23OAuth 2.0 Token Exchange (RFC 8693)↗ rfc-editor.org

  Delegated authorization patterns, sub (subject) vs. act (actor) claims for agent delegation.

• R37SLSA & Sigstore↗ slsa.dev

  Supply-chain Levels for Software Artifacts (SLSA), build provenance attestations, and Cosign artifact signing.

Read only the named sections needed for today's task. Reading sits inside the 60-minute concepts block; it is not extra time.

• docsSLSA (OpenSSF)
  Security levels

  Defines build levels L0 to L3 and what signed provenance from a hardened platform proves, which is the attestation baseline for pinning and verifying model artifacts in D12.

  R37checked 2026-11-11
• videoDaily AI News By Ali · 8:29
  Find Leaked API Keys Before Anyone Else: Free Gitleaks Tutorial in 5 Steps

  Runs a gitleaks scan across a repository so you can reproduce the detection half of the compromised-credential drill before rotating anything.

  R22checked 2026-11-11
• videoIBM Technology · 5:24
  Building Trustworthy AI: Avoid Model Drift & Unsafe Outputs

  Explains detecting behaviour drift and unsafe outputs in deployed models, the signal that a silently altered checkpoint or provider update has shipped.

  R11checked 2026-11-11
• articleHugging Face
  Model Cards

  Documents the metadata that pins a model to a named base, dataset and license, the fields you diff to prove the deployed artifact matches the approved one.

  R11checked 2026-11-11
• repoGitHub
  gitleaks/gitleaks

  The scanner itself, with git-history, directory and stdin modes plus a baseline flag, so you can find the leaked key across the whole commit history immediately after revocation.

  R22checked 2026-11-11

A lab is complete only when it produces evidence. Installing PyRIT, Garak, Presidio or TransformerLens is not completion. Saved on this device only.