How a submission becomes verified truth.
No ranger writes their own score. The program measures each repository, signs the result, and this site proves the signature. Here is the exact path, the identity model, and the gate.
The pipeline
Day work
Reproduce and fix in the loopback lab; capture a negative control.
evidence/W##/D###
README + results file, placeholders resolved, attested no-AI.
Pull request
One per day, branch day/Dxxx, protected main.
test-and-validate
pytest, ownership check, loopback + secret guard. Red blocks merge.
Signed release
A board turns green only on tag board1..6.
Program re-measures
Program CI re-runs the evidence scan per repo, then Ed25519-signs the cohort.
Verified in-browser
Your browser checks the signature before any number renders.
Who we trust, and why it cannot be faked
The program measures, not the ranger
Cohort sync clones each rostered repo and re-runs the same evidence scan. A ranger editing their own local status changes nothing: the authoritative number is recomputed by the program.
One signature over the whole cohort
The aggregate is signed with an Ed25519 key held only as a CI secret. The public key ships with the site; the private key never leaves the program's CI, so only the program can produce a valid board.
Tamper is visible, not silent
The signature covers the exact bytes served. Change one digit of one percentage and every visitor's browser shows "signature invalid" instead of trusting the data.
# program CI, per sync (scheduled + on demand) for repo in roster: git clone --depth=1 $repo _repos/$handle python3 scripts/aggregate_cohort.py --repos-dir _repos \ --out portfolio/site/src/data/cohort.json # sign with the CI secret, never a committed key COHORT_SIGNING_KEY=$SECRET python3 scripts/sign_cohort.py \ portfolio/site/src/data/cohort.json # Vercel redeploys; the site verifies the signature on load
How red/blue engagements are scored
Every engagement is a pair: a red action and the blue detection meant to catch it, mapped to a MITRE ATT&CK technique. The program records the outcome and the time to detect from the ranger's own SIEM and detection fixtures.
Outcomes
What counts
A detection passes only with a positive fixture that fires and a negative fixture that stays quiet. Mean time to detect is measured from the red action's first event. Missed engagements drive the next detection-engineering task; nothing is scored on a screenshot.