assumption-excavation
Surface hidden assumptions in plans, specs, AGENTS.md, SKILL.md, SET bundles, and workflow contracts before implementation or handoff. Use when an artifact looks plausible but may rely on unstated environment, dependency, behavioral, temporal, or success assumptions.
Install
npx skills add https://github.com/markoblogo/abvx-agent-skills/tree/main/skills/assumption-excavation
claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install markoblogo-abvx-agent-skills@llmmart
git clone https://github.com/markoblogo/abvx-agent-skills.git
The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole markoblogo/abvx-agent-skills collection as a plugin from our marketplace. Git is the plain clone.
Skill manifest
Assumption Excavation
Find what the artifact takes for granted before those assumptions become implementation bugs.
Use When
- reviewing PRDs, specs, plans, or issue slices before work starts;
- reviewing
AGENTS.md,SKILL.md, SET bundles, proof-loop contracts, or workflow docs; - a proposal sounds confident but depends on unclear prerequisites;
- you need a lightweight pre-implementation review lens.
Workflow
- Identify the artifact and its intended consumer.
- Read for implicit assumptions, not prose quality.
- Classify assumptions:
- environment: tools, services, files, auth, OS, network;
- dependency: upstream state, input shape, prior checks;
- behavioral: what humans, agents, or systems are expected to do;
- temporal: what must remain stable over time;
- success: what the artifact counts as done or safe.
- Rank the assumptions by fragility:
- high: likely to break or expensive if wrong;
- medium: plausible risk with manageable fallout;
- low: visible but not blocking.
- Mark partially surfaced assumptions as
PARTIAL. - Recommend what must be clarified, tested, or accepted before implementation.
Output Shape
Use:
- verdict:
EXAMINEDorUNEXAMINED; - top assumptions, ranked;
- fragility score or severity per item;
- missing evidence;
- smallest clarification or verification step.
Guardrails
- Do not rewrite the artifact unless asked.
- Do not treat assumptions as errors by default; some are deliberate.
- Do not block on low-risk assumptions.
- State uncertainty as text-based inference, not mind reading.
Provenance
Adapted from the assumption-excavator pattern in aself101/agents-and-pipelines, compressed for ABVX skill use and broader repo-contract review.
Files (abvx-agent-skills)
-
agents
-
openai.yaml 249 B
Loading…
-
-
SKILL.md 2.1 KB
--- name: assumption-excavation description: Surface hidden assumptions in plans, specs, AGENTS.md, SKILL.md, SET bundles, and workflow contracts before implementation or handoff. Use when an artifact looks plausible but may rely on unstated environment, dependency, behavioral, temporal, or success assumptions. license: MIT metadata: abvx_status: experimental abvx_origin: adapted --- # Assumption Excavation Find what the artifact takes for granted before those assumptions become implementation bugs. ## Use When - reviewing PRDs, specs, plans, or issue slices before work starts; - reviewing `AGENTS.md`, `SKILL.md`, SET bundles, proof-loop contracts, or workflow docs; - a proposal sounds confident but depends on unclear prerequisites; - you need a lightweight pre-implementation review lens. ## Workflow 1. Identify the artifact and its intended consumer. 2. Read for implicit assumptions, not prose quality. 3. Classify assumptions: - environment: tools, services, files, auth, OS, network; - dependency: upstream state, input shape, prior checks; - behavioral: what humans, agents, or systems are expected to do; - temporal: what must remain stable over time; - success: what the artifact counts as done or safe. 4. Rank the assumptions by fragility: - high: likely to break or expensive if wrong; - medium: plausible risk with manageable fallout; - low: visible but not blocking. 5. Mark partially surfaced assumptions as `PARTIAL`. 6. Recommend what must be clarified, tested, or accepted before implementation. ## Output Shape Use: - verdict: `EXAMINED` or `UNEXAMINED`; - top assumptions, ranked; - fragility score or severity per item; - missing evidence; - smallest clarification or verification step. ## Guardrails - Do not rewrite the artifact unless asked. - Do not treat assumptions as errors by default; some are deliberate. - Do not block on low-risk assumptions. - State uncertainty as text-based inference, not mind reading. ## Provenance Adapted from the `assumption-excavator` pattern in `aself101/agents-and-pipelines`, compressed for ABVX skill use and broader repo-contract review. -
SKILL_CARD.md 1.4 KB
Loading…
Comments (0)
Sign in to join the conversation.
Reviews (0)
No reviews yet.
No comments yet.