Anomaly Detection
Define behavioral baselines and anomaly detection rules for "$ARGUMENTS". Establish normal patterns, create detection rules for deviations, and configure automated responses for high-confidence anomalies.
Arguments. $ARGUMENTS may contain the project name and optional flags. Take the first token as <project> and treat --skip-checks as a flag; never include flags in file paths. All paths below use <project>. If --skip-checks is present, skip the prerequisite check and, before writing any artifact, append one line to .singularskills/log.jsonl: {"ts":"<ISO timestamp>","skill":"security-anomaly-detection","workflow":"security","phase":3,"project":"<project>","event":"skipped","artifact":"03-detect.md","note":"<reason or \"none given\">"}.
Prerequisites
Part of a workflow. This skill is phase 3 (Detect) of the NIST CSF workflow (security); cross-domain workflows run it too. First check .singularskills/log.jsonl for an open run for this project: a run-started line with "project":"<project>" and no later run-finished line for the same workflow and project. If one exists, the conductor is running you: continue here. Otherwise, if the user asked to start, run, continue or resume a process or workflow rather than this one step, do not continue here: invoke the workflow skill with the workflow the user asked for, or security when they named none, followed by <project>; it runs this phase in order, asks before it writes and records the run. Otherwise continue here: a later phase runs on its own, behind its gate.
Run this skill's gate before writing: sh <this skill's folder>/scripts/gate.sh <project> <project root>, where <project root> is the folder that contains .singularskills/ (use . when you are already there). If it exits non-zero, stop and tell the user what it printed, unless --skip-checks was given (see Arguments). On a refusal, before you stop, append this line to .singularskills/log.jsonl (create the folder and file if missing): {"ts":"<ISO timestamp>","skill":"security-anomaly-detection","workflow":"security","phase":3,"project":"<project>","event":"refused","artifact":"03-detect.md","note":"<what the gate printed>"}. Once the gate has passed, or under --skip-checks once you have appended the skipped line, append this line before you start work: {"ts":"<ISO timestamp>","skill":"security-anomaly-detection","workflow":"security","phase":3,"project":"<project>","event":"started","artifact":"03-detect.md","note":""}.
Read .singularskills/security/<project>/00-govern.md. If this file does not exist, tell the user:
Phase 0 (Govern) has not been completed for "". Run a Govern skill first (e.g.,
/security-security-policy <project>), or use--skip-checksto bypass.
Process
Establish behavioral baselines:
- Login patterns — normal login times, locations, devices, and frequency per user population
- API usage patterns — typical request rates, endpoint access distribution, payload sizes
- Data access patterns — normal query volumes, data export sizes, access to sensitive resources
- Network patterns — expected traffic volumes, destination IPs/domains, protocol distribution
- System patterns — normal CPU/memory usage, process lists, file system activity
- Define baseline measurement period (minimum 30 days of data, accounting for business cycles)
Define anomaly detection rules:
- Impossible travel — login from geographically distant locations within an impossible timeframe
- Unusual hours — access outside established working hours for the user's role and timezone
- Failed login spike — sudden increase in failed authentication attempts (brute force indicator)
- Bulk data export — data retrieval significantly exceeding baseline (exfiltration indicator)
- Privilege escalation attempts — repeated access denied events followed by elevated access
- New admin accounts — creation of privileged accounts outside change management process
- Lateral movement — sequential access across multiple systems from a single compromised account
- DNS anomalies — queries to newly registered domains, high-entropy domain names (DGA detection)
Set detection thresholds:
- Define statistical thresholds (standard deviations from baseline, percentile-based)
- Tune thresholds to minimize false positives while maintaining detection sensitivity
- Implement different thresholds by user population (admins vs. standard users, humans vs. service accounts)
- Define learning period for new users and systems before enforcing anomaly rules
- Plan for seasonal adjustments (end-of-quarter, product launches, maintenance windows)
Create alert correlation rules:
- Combine multiple low-confidence signals into high-confidence detections
- Define attack chain correlations (reconnaissance -> initial access -> lateral movement -> exfiltration)
- Implement time-windowed correlation (related events within a defined time period)
- Create entity-based correlation (aggregate anomalies per user, per IP, per service)
- Assign composite risk scores based on correlated signals
Define automated response actions:
- High confidence — automatic containment (account lockout, session termination, network isolation)
- Medium confidence — step-up authentication (require MFA re-verification), increase logging verbosity
- Low confidence — flag for analyst review, add to watchlist for enhanced monitoring
- Define rollback procedures if automated response was a false positive
- Ensure all automated actions generate audit trail entries
Write the artifact — append to
.singularskills/security/<project>/03-detect.mdunder a## Anomaly Detectionsection. Create the file with a# Detect — <project>heading if it does not exist; if a## Anomaly Detectionsection already exists, replace only that section. The section contains:- Behavioral Baselines — patterns measured, data sources, and measurement methodology
- Detection Rules — each anomaly rule with logic, data source, and threshold
- Threshold Configuration — statistical methods, tuning approach, and false positive targets
- Correlation Rules — multi-signal correlations and composite risk scoring
- Automated Responses — actions by confidence level with rollback procedures
- Tuning Plan — ongoing threshold adjustment and false positive reduction process
Output
The anomaly detection plan appended to .singularskills/security/<project>/03-detect.md. Present a summary to the user highlighting:
- Behavioral baselines to be established and data sources
- Key anomaly detection rules and their threat coverage
- Automated response actions by confidence level
- Correlation rules for attack chain detection
Record the run. After you have written your section, append this line to .singularskills/log.jsonl (the started line of Prerequisites, with "event":"finished"): {"ts":"<ISO timestamp>","skill":"security-anomaly-detection","workflow":"security","phase":3,"project":"<project>","event":"finished","artifact":"03-detect.md","note":""}.
This artifact is shared by every skill of phase 3 (Detect). The frontmatter is written once, by whoever creates the file: if you create it, open it with the block below (skill is the phase title, started the time of that first write; gate is passed when the gate ran clean and skipped under --skip-checks); if it already exists, leave the block alone except to update finished and to set gate to the worst value among the sections (skipped over passed).
---
skill: Detect
workflow: security
phase: 3
project: <project>
started: <ISO timestamp>
finished: <ISO timestamp>
agent: <the agent you run in, or "unknown">
gate: <passed|skipped>
---
Open your ## Anomaly Detection section with one italic line, *<skill name> · <ISO timestamp> · gate: <passed|skipped>* (your skill name is security-anomaly-detection).
End your section with ### Decisions and ### Open questions (one level under your ## Anomaly Detection heading, so the section stays one block), in this shape:
### Decisions
- <what was chosen> — <why> — <what was rejected>
### Open questions
- <what the next phase or a reviewer must settle>
End the artifact with this paragraph, verbatim:
Not a penetration test or audit. This artifact is generated analysis; validate findings against the live system and have material controls reviewed by a qualified security professional.