Liberty91LTD/cti-skills / skills/quality-of-information-check
quality-of-information-check
VerifiedUse when evidence needs grading before anyone reasons from it: "how reliable is this?", "grade this source", "check this report", "is this corroborated?", or a URL, pasted report or Liberty91 event handed over ahead of an ACH or assessment. Resolves each document to its originating source, splits it into claims and grades each on access level and claim support with a rationale and hard caps; not an Admiralty rating. Returns a graded claim table, gaps, a deception screen and JSON. Run first by /ach, /threat-assessment and /writing-assessments.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 15.9 KB |
| references/aliases.example.yml | config | 1.7 KB |
| references/deception-screen.md | markdown | 5.6 KB |
| references/doctrine.md | markdown | 6.6 KB |
| references/evidence-item-schema.md | markdown | 10.2 KB |
| references/grading-rubric.md | markdown | 12.2 KB |
| references/worked-examples.md | markdown | 11.3 KB |
| scripts/validate_evidence.py | script | 25.2 KB |
references/deception-screen.md
View on GitHub# Deception screen
Four checks from Heuer and Pherson's deception detection technique, with prompts written for CTI source material. The screen inside a Quality of Information Check is short: a few sentences per check. It raises flags. It does not settle whether deception occurred.
Run it in full when:
- any claim is an `actor_claim`;
- attribution rests on artefacts an adversary controls (language strings, compile times, tool choice, infrastructure reuse);
- a single source carries the bottom line;
- the information arrived at a moment that suits someone.
Answer from the documents and from resolved data. Where a prompt asks about history ("does this actor inflate its claims?"), use a knowledge cell or a lookup and cite it, or answer "not established".
## MOM: motive, opportunity, means
Could someone be deceiving us, and would they want to?
| Prompt | Notes |
|---|---|
| Who benefits if this claim is believed? | The actor (pressure on the victim, reputation among affiliates), the victim (minimising), the vendor (commercial interest), a state (deflection) |
| Does the actor gain from overstating scale? | Extortion leverage depends on the victim and the public believing the figure |
| Does the victim gain from understating? | Disclosure wording is drafted with legal and market exposure in mind |
| Does the primary have a commercial interest in this finding? | Read `primary_source.interest`. A vendor reporting on a threat its product addresses, or on a competitor's product, sets `commercial_interest` |
| Could the adversary have planted the artefacts the attribution rests on? | Language settings, code comments, reused tooling and working-hours patterns are all within the adversary's control |
| Did the source have the opportunity to know what it claims? | This is the access question again. An actor knows what it took. A vendor without an engagement at the victim does not |
## POP: past opposition practices
Has this adversary deceived before, and how?
| Prompt | Notes |
|---|---|
| Does this actor have a history of inflating victim counts or data volumes? | Check `/ransomware-ecosystem`, `/lookup-ransomwarelive` |
| Has this actor re-posted old or third-party data as a new breach? | A recurring pattern on leak sites and forums |
| Has this actor or its sponsor used false-flag techniques? | Check the relevant regional knowledge cell |
| Is this a persona with a record of claiming others' operations? | Common among hacktivist fronts. Check `/hacktivism` |
| Does the claim fit the pattern of the actor's earlier statements that later proved accurate or inaccurate? | If no record is available, say so |
## MOSES: manipulability of sources
How easily could the source have been fed this?
| Prompt | Notes |
|---|---|
| Does the claim rest on a single sample, a single upload, or a single forum post? | Anyone can upload a file to a public sandbox |
| Is the source relaying what it was told, without its own observation? | A journalist quoting the actor, a vendor quoting a leak site |
| Could the source be the target of the message? | Actors speak to journalists and researchers because they will repeat it |
| Does the chain pass through a platform where identity is unverified? | Social posts and Telegram channels. The account holder is the source, and the account may not be who it says |
| Is the indicator set something an adversary could seed? | Infrastructure that is abandoned, shared or deliberately exposed |
## EVE: evaluation of evidence
Does the evidence hold together?
| Prompt | Notes |
|---|---|
| Is the claim consistent with the other claims from the same primary? | Internal contradictions lower claim support to disputed |
| Is it consistent with other primaries? | A conflict between two credible primaries sets `disputed` |
| Do the dates work? | Activity dated after publication, a "new" sample with an old compile time and old submissions |
| Is anything conspicuously missing? | Attribution with no stated basis. A breach claim with no sample data. An exploitation claim with no indicators |
| Is the evidence too neat? | Every artefact pointing at one actor, with no noise, in a domain where artefacts are cheap to forge |
| Has any part been retracted, corrected or superseded? | Set `retracted` or `superseded` |
## Recording the result
In the human-readable output, four short paragraphs or a four-row table, then the flags raised.
In the JSON envelope:
```json
"deception_screen": {
"mom": "The 400GB figure serves the actor's extortion leverage. The vendor has a commercial interest but reports no finding that depends on its own product.",
"pop": "Not established. No record of this actor's earlier claims was available in this run.",
"moses": "C07 rests on a single leak-site posting relayed by the press.",
"eve": "Dates are consistent. The attribution states its basis. No sample data accompanies the volume claim.",
"flags_raised": [
{"claim_id": "C07", "flag": "deception_indicators_present", "reason": "Actor has motive to inflate; single unverified posting; no sample data."}
]
}
```
A raised flag does not change a grade by itself. It is recorded on the claim, shown to the reader, and carried into `/ach`, where a hypothesis that depends on a flagged claim is marked as doing so.
## What the screen is not
It is not the full deception detection technique, which is a separate structured exercise. And it is not a reason to discount everything an actor says. Actors' claims are often true. Their access level is adversary because the source has a motive to mislead and cannot be held to account, and their claim support rises when someone with separate access confirms them.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |