source-analysis

name	source-analysis
description	Evaluate a single source across 7 structured fields (source type, authority, methodology, evidence quality, limitations, bias, bottom line) with an explicit citation recommendation.
tools	["claude-code"]
trigger_examples	["Evaluate / critique / analyze this paper","How good is this source?","Can I cite this?","I'm thinking of citing [X] for [claim]"]

When to use

Use this skill when the user has a single source — a paper, article, book chapter, or preprint — and needs a structured evaluation of its credibility, authority, and citability. Source analysis produces a judgment, not a summary.

Distinct from literature review and research synthesis: those skills work across multiple sources. Source analysis focuses entirely on one source and delivers a citation recommendation as its final output. Use it when the key question is "Can I cite this? For what kinds of claims? With what caveats?"

If the user shares multiple sources and wants comparison, prefer literature review or research synthesis instead.

Triggers

• User says "Evaluate / critique / analyze this paper" or "How good is this source?" — highest-confidence direct invocation.
• User pastes a single paper abstract or excerpt with no specific ask — proactively offer source analysis.
• User asks "Can I cite this?" or "Is this source any good?" — offer source analysis with citation-recommendation framing.
• User says "I'm thinking of citing [X] for [claim]" — offer source analysis with Bottom line tuned to the specific claim.

Instructions

1. Read the provided source fully before beginning the evaluation.
2. State the source type at the top of the output (Primary / Secondary / Tertiary) with a one-sentence justification so the user can recalibrate citation strategy before reading further.
3. Complete all 7 fields in order. Do not skip or mark a field "N/A" without explanation.
4. Source type — Primary (original empirical data or argument), Secondary (review, synthesis, or commentary on primary sources), or Tertiary (meta-review of reviews, encyclopedias, textbooks). Justify in one sentence.
5. Authority — author credentials, institutional affiliation, track record (h-index or key prior work if relevant), and domain fit. Note if the author is writing outside their primary field of expertise.
6. Methodology — study design (empirical / theoretical / review / meta-analysis), sample, and measures where applicable. Assess whether the method is appropriate for the main claim.
7. Evidence quality — strength of evidence supporting the main claim. Does it converge with other studies or is it an isolated result? Note replication status if known.
8. Limitations — distinguish what the author acknowledges and what the author omits. Every source has limitations; a complete analysis names both categories.
9. Bias assessment — funding source, ideological framing, methodological narrowness, and motivated framing. State bias level explicitly (low / moderate / high) with reasoning.
10. Bottom line — write a narrative judgment paragraph answering: Can this source be cited? For what kinds of claims? With what caveats? Apply the writing-profile rule from ## Writing-profile integration to this paragraph.
11. Apply the BibTeX extension if the user supplies citation keys: use @key references throughout.

Output format

Plain GitHub-flavored markdown in the chat.

Source type declaration (required, at top): **Source type:** [Primary/Secondary/Tertiary] — one-sentence justification.

7 fields as labeled list: Each field rendered as a **Field name:** bold-label line, with field content on the same or following line. Fields 1–6 are typically 1–3 sentences each. Use complete sentences, not bullets, within each field.

Bottom line: A standalone paragraph (typically 100 words or more) that delivers the citation recommendation. This is the interpretive payoff of the analysis.

No tables by default — the 7 fields are a fixed schema, not a comparison matrix. No Obsidian [[wikilinks]]. No JSON or YAML sidecar. Portable across academic writing tools and plain-text editors.

Quality criteria

1. Source type (primary/secondary/tertiary) stated explicitly at top with one-sentence justification.
2. All 7 fields populated — no field skipped or marked "N/A" without explanation.
3. Authority field addresses credentials, domain fit, and track record — not just the author's name.
4. Methodology field describes design, sample, and measures where applicable.
5. Limitations section distinguishes author-acknowledged from author-omitted limitations.
6. Bias assessment is explicit (low / moderate / high) with reasoning.
7. Bottom line delivers a citation recommendation — answers "can I cite this? for what? with what caveats?"

Anti-patterns

• Summarizing the source instead of evaluating it — a good summary is not source analysis.
• Accepting the author's framing uncritically — analysis must surface what the author is not saying.
• Ignoring methodology — citing a 30-subject pilot and a 3000-subject RCT as equivalently authoritative.
• No authority assessment — failing to note when a source is outside the author's field of expertise.
• No limitations surfaced — every source has limitations; if none are listed, the analysis is incomplete.
• Conflating primary with secondary — treating a review paper's claim as though it were original evidence.
• No explicit bottom-line judgment — analysis must end with a citation recommendation, not drift into summary.

Example

Input: Baddeley (2000). "The Episodic Buffer: A new component of working memory?" Trends in Cognitive Sciences, 4(11), 417–423.

Output:

# Source analysis: Baddeley (2000) — The Episodic Buffer

**Source type:** Secondary (theoretical review article extending a prior model).

**Authority:** Alan Baddeley — emeritus professor, University of York; co-originator of the working-memory model (Baddeley & Hitch 1974); h-index >130; this paper is within the author's primary research domain and extends his own theoretical framework.

**Methodology:** Theoretical paper — no new empirical data. Synthesizes prior empirical findings (cross-modal binding, prose recall, amnesiac integration performance) to motivate a new model component. Argument structure: (1) limits of the original 3-component model, (2) empirical anomalies the buffer resolves, (3) buffer properties.

**Evidence quality:** Claims are consistent with the cited empirical record at the time. The episodic buffer proposal has since been adopted as standard in the working-memory literature (see Baddeley 2012 for follow-up). However, direct empirical evidence for the buffer as a structurally distinct component (vs. an emergent property of executive attention) remains contested.

**Limitations acknowledged by author:** "The nature of the buffer is still to be determined" — Baddeley notes the proposal is a placeholder for integrating capacity. Limits on buffer capacity and its relationship to long-term memory are left as open questions.

**Limitations omitted by author:** The paper does not engage seriously with attentional-focus alternatives (Cowan 1999) that can account for the same empirical anomalies without requiring a new structural component. The buffer is presented as the resolution; alternative models are mentioned only briefly.

**Bias assessment:** Low-to-moderate. The author is extending his own model — mild theoretical commitment bias. No funding-source concerns. Publication venue (*Trends in Cognitive Sciences*) is reputable and peer-reviewed.

**Bottom line:** Citable as the canonical statement of the episodic buffer proposal. Appropriate for: (1) describing the 4-component working-memory model in its mature form, (2) grounding discussions of binding and integration in working memory. Caveat: when citing, acknowledge the buffer remains theoretically contested and Cowan's attentional-focus alternative should be mentioned in any thorough treatment.

Writing-profile integration

Two-tier rule based on output section:

• Fields 1–6 (Source type, Authority, Methodology, Evidence quality, Limitations, Bias assessment): structured judgment — terse clauses, minimal prose. context/writing-profile.md is not consulted; these fields are structured data.
• Bottom line: Narrative judgment paragraph — full writing-profile consultation applies. Tone, sentence-length preferences, and pet peeves from context/writing-profile.md apply to this paragraph only. The Bottom line typically exceeds 100 words, triggering the adapt-to-profile rule.

Example prompts

• "Analyze this paper: [paste abstract or filename]."
• "Is this source credible and relevant to my research on [topic]?"
• "Can I cite this for my claim about [X]?"
• "What are the methodological strengths and limitations of this study?"