assess-assignment

name	assess-assignment
description	Assessment guide for evaluating Build on Andamio module assignments. Used by learn or invoked directly when a learner submits evidence. Covers how to evaluate evidence against SLT criteria, apply rubrics, and provide constructive feedback.
license	MIT
metadata	{"author":"Andamio","version":"0.2.0"}

Skill: Assess Assignment

Evaluate a "Build on Andamio" module assignment. This skill is agent-neutral: use it inline when subagents are unavailable.

Path Resolution

• Plugin context (${CLAUDE_PLUGIN_ROOT} is set): read course content from ${CLAUDE_PLUGIN_ROOT}/courses/.
• Clone/symlink context: read course content from courses/ relative to the project root.

Assessment Protocol

Step 1: Read The Assignment File

Read courses/build-on-andamio/lessons/m{N}/assignment.md for the module being assessed.

This file defines:

• evidence required for each SLT
• what a strong answer looks like
• expected format, such as written response, CLI output, transaction hash, or address

Also read courses/build-on-andamio/01-slts.md for canonical SLT wording.

Step 2: Evaluate Each SLT

For written responses:

• check whether the answer demonstrates the capability named by the SLT
• require the specific elements the assignment says a strong answer should include
• accept different phrasing when the reasoning is sound
• ask for revision if the answer only parrots lesson text without showing understanding

For CLI output:

• check that output is present and non-empty
• check that the format plausibly matches the command
• for version/status checks, any valid non-error output is usually enough

For transaction hashes and addresses:

• transaction hashes should be 64 hex characters
• preprod addresses should start with addr_test1
• if verification is requested and Andamioscan is reachable, verify against preprod.andamioscan.andamio.io
• do not fail solely because Andamioscan is unreachable when format evidence is otherwise plausible

For endpoint/spec lookups:

• check that the learner found the right endpoint
• require a short explanation of what it does and what parameters it takes

Step 3: Assess Feedback Section

Part 2 of every assignment is feedback on the module. It is required but not graded.

• check that it is present and non-empty
• acknowledge substantive feedback specifically
• accept simple feedback such as "everything was fine"
• note anything actionable for course improvement

Step 4: Determine Verdicts

Per-SLT verdicts:

Verdict	Criteria
Pass	Evidence is present and demonstrates the SLT capability. Minor phrasing gaps are acceptable.
Revise	Evidence is present but incomplete, vague, partially incorrect, or wrong format.
Missing	No evidence provided for the SLT.

Module verdict:

Verdict	Criteria
Accept	All SLTs pass and the feedback section is present.
Revise	One or more SLTs are Revise or Missing.

Step 5: Write The Assessment

Use this structure:

## Module {N} Assessment

### SLT {N}.1 — {title}
**Verdict:** Pass | Revise | Missing
{2-3 sentences: what was strong, what needs work. Be specific.}

### SLT {N}.2 — {title}
**Verdict:** Pass | Revise | Missing
{2-3 sentences}

### Feedback Section
{Acknowledge the learner's feedback. Note anything actionable.}

### Module Verdict: Accept | Revise
{If Accept: name what the learner can now do.}
{If Revise: list exactly which SLTs need revision and what is missing.}

Calibration Notes

Be the reviewer described in M100.1: a domain reviewer who can tell whether the work is real. Credentials are meaningful only if the review is credible.

Common revision patterns:

• paraphrased lesson with no learner reasoning: Revise
• tx hash without required address, or address without required tx hash: Revise for the missing item
• correct but very brief written response: Pass only if the key insight is clearly present
• "I don't know" for an SLT: Missing

For resubmissions, focus on whether the prior gaps were addressed. Do not re-evaluate already-passed SLTs unless the learner changed the answer.

Module-Specific Notes

M100 Foundations

• SLT 100.1 is the philosophical anchor. Look for genuine engagement with the three learning models.
• SLT 100.2 and 100.3 are simple verification items.
• SLT 100.4 asks about the relationship between understanding and ownership. Look for the learner's own insight.

M200 Identity

• SLT 200.1 has multiple evidence items. Do not pass if one is missing.
• SLT 200.3 is a status check. Any non-error output can pass.
• SLT 200.4 is conceptually tricky. The Attestation JWT vs User JWT distinction should be crisp.

M300 Reading Andamio

• Assess navigation and interpretation, not memorization.
• Accept any reasonable method of finding the right endpoint.
• Hold a high bar for interpreting the source field.

M400 TX State Machine

• SLT 400.2 should explain the timing gap that makes updated safer than earlier statuses.
• SLT 400.4 has multiple scenarios; all should be addressed.
• SLT 400.5 has a precise answer; both reverting states and reasons should be present.