decontextualize
Decontextualize
You're overfitting to the current conversation. The artifact you're creating will outlive this context.
The Problem
When you have rich context (current task, domain jargon, specific examples), you default to using it—even when the artifact should be portable. You optimize for the current conversation instead of the future reader.
Symptoms
- Examples drawn from current work ("Add validation to the user signup flow")
- Domain jargon the reader won't know
- Phrasing that mirrors the user's exact words instead of standing alone
- References that require conversation history to understand
The Fix
1. Identify the artifact's audience
| Artifact | Audience | Context they have |
|---|---|---|
| Commit message | Repo contributors | Repo history |
| PR description | Reviewers | Repo + diff |
| Config/rules file | Future self, strangers | None |
| Library docs | External users | None |
| Skill/prompt | Any project | None |
2. Apply the stranger test
"Would someone with zero context about this conversation understand this?"
If any example, term, or reference requires conversation history—replace it.
3. Genericize examples
| Before (context-bound) | After (portable) |
|---|---|
| "Fix the auth bug in UserService" | "Fix authentication timeout" |
| "Update the Client class" | "Update the API client" |
| "like we discussed" | [remove or specify] |
4. Check your language
- Don't parrot the user's exact phrasing if it's overly specific
- Use terminology that's standard in the domain, not coined in-conversation
- Write for the stranger who arrives later
On Activation
- Identify which artifact is overfitting
- List the context-bound elements
- Propose generic replacements
- Apply the fix
No preamble. Just fix it.
More from corygabrielsen/skills
mission-control
Coordinate complex multi-step work using task graphs and parallel background agents. Use when work requires decomposition, delegation, and long-running operations that may survive context compaction.
11loop-codex-review
Automated code review loop with progressive reasoning levels. Runs n parallel Codex reviews (configurable via -n), Claude addresses issues, climbs from low→xhigh reasoning until fixed point (all n clean). Human approval at each iteration.
11review-pr
Thorough, educational PR review process. Gathers context, explains changes, identifies issues systematically, and drafts constructive feedback collaboratively before submitting.
11orthogonalize-pr
Identify and separate orthogonal (independent) change sets within a decomposed PR. Creates branches for each set that could theoretically be reviewed or merged separately.
10loop-review-skill-until-fixed-point
Iterate /review-skill on a target until fixed point. Runs review passes until all reviewers return NO ISSUES.
9fork
Branch off a conversation to handle tangents. Outputs context summary and ready-to-paste command for a new terminal session.
9