agentleFS
davidmatousekdavidmatousek/tachi/.claude/skills/tachi-control-analysis/SKILL.md

tachi-control-analysis

Skill95 starsChanged 37 days agoApache-2.0View on GitHub

Domain knowledge for compensating controls analysis — control category definitions with detection patterns, evidence criteria with effectiveness classification, and residual risk calculation with recommendation generation. Loaded on-demand by the control-analyzer agent during codebase scanning and risk assessment phases.

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When baseline controls are available:

    line 54
  2. Very specific:

    When no baseline compensating-controls.md exists, all findings are scanned with rescan_scope: "full" and control_carry_forward: false.

    line 68

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Tachi Control Analysis Skill

This skill contains the domain knowledge extracted from the tachi control-analyzer agent. It provides the reference data needed to detect compensating controls in a target codebase, classify their effectiveness, calculate residual risk, and generate remediation recommendations.

Domain Overview

The control analysis domain covers three areas:

  1. Control Categories and Detection Patterns -- Definitions for the 8 compensating control categories (authentication, input-validation, rate-limiting, encryption, logging-audit, csrf-protection, csp-security-headers, access-control), their STRIDE-to-control mapping, pattern indicators for Phase A scanning, and common library/framework references.
  2. Evidence Criteria and Effectiveness Classification -- Phase B semantic analysis criteria (context checks, enforcement checks, strength assessments), evidence collection rules (snippet selection, deduplication, file path format), confidence level definitions (High/Medium/Low), and Phase 4 classification rules (found/partial/missing with multi-control resolution and cross-component handling).
  3. Residual Risk Calculation and Recommendations -- Recommendation generation rules for missing and partial controls (templates, effort calibration), residual score computation formula with the P0 binary reduction model (reduction factors by control status), severity band mapping for residual scores, and summary statistics calculations.

NIST AI RMF Relationship

Tachi's compensating-controls analyzer operates in a STRIDE+AI idiom, not a NIST idiom. The NIST AI Risk Management Framework (AI RMF 1.0, NIST AI 100-1) and its companion Generative AI Profile (NIST AI 600-1) are the U.S. federal reference vocabulary for managing AI system risk — Functions (Govern, Map, Measure, Manage), Subcategories, and 12 GAI risk categories. Tachi's posture toward both documents is documentation-only mapping per ADR-025: no schema field, no agent, no pipeline phase, and no SARIF tag emits NIST-keyed output. The strongest direct semantic overlap is MEASURE 2.7 ("AI system security and resilience are evaluated and documented") — essentially what compensating-controls.md already produces without NIST labeling. Adopters who must cite NIST mappings during procurement, audit, or examination workflows should consult the companion reference .claude/skills/tachi-shared/references/nist-ai-rmf-mapping.md and ADR-025 for the three-surface comparison (Functions × phases, Subcategories × control categories, GAI risks × STRIDE+AI), the full re-evaluation triggers, and the structural-fit rationale for choosing mapping over wired integration.

Baseline-Aware Control Analysis Rules

Carry-Forward Conditions

Control status is carried forward from baseline when ALL of the following are true:

  1. Finding delta_status is UNCHANGED
  2. Baseline compensating-controls.md exists and is parseable
  3. Finding ID matches a baseline controlled finding (exact findingId/v1 match)

Carry-Forward Fields

For UNCHANGED findings, copy these fields verbatim from baseline:

FieldTreatment
control_statusCopy (found/partial/missing)
control_evidenceCopy all evidence entries
control_categoryCopy
control_effectivenessCopy
reduction_factorCopy
residual_scoreCopy
residual_severity_bandCopy
recommendationCopy (if any)
effort_estimateCopy (if any)

Set control_carry_forward: true and rescan_scope: "incremental".

Incremental Re-Scan Scope

When baseline controls are available:

ScenarioRe-Scan ScopeRationale
All findings UNCHANGEDincremental (nothing to scan)Full inheritance
Mix of UNCHANGED + NEW/UPDATEDincremental (scan only changed)Partial inheritance
All findings NEW (first run or no baseline controls)fullNo inheritance possible

Evidence Preservation

Carried-forward evidence entries retain their original file paths, line numbers, and snippets from the baseline run. These represent the control state as of the baseline — they are not re-validated against the current codebase for UNCHANGED findings.

First Run Behavior

When no baseline compensating-controls.md exists, all findings are scanned with rescan_scope: "full" and control_carry_forward: false. No behavioral change from pre-baseline pipeline.

Reference Loading Table

Load reference files on-demand using the Read tool at the workflow phase where they are needed. Do not load all references at pipeline start.

Reference FileLoad WhenContent
references/control-categories.mdPhase 3 (Detect Controls)8 control category definitions, STRIDE-to-control mapping, pattern indicators, common libraries
references/evidence-criteria.mdPhase 3 (Detect Controls), Phase 4 (Map & Classify)Evidence collection rules, confidence levels, classification rules (found/partial/missing), multi-control resolution
references/residual-risk.mdPhase 5 (Recommend & Calculate Residual Risk)Recommendation templates, effort calibration, residual score formula, reduction factors, severity band mapping, summary statistics

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.