agentleFS
MicrosoftKnown publishermicrosoft/tlalink/.github/skills/tlalink-source-navigation/SKILL.md

tlalink-source-navigation

Skill1 starsChanged 14 days agoMITView on GitHub

Find C# symbols and Roslyn declaration IDs with TlaLink; resolve source positions, inspect member code, find references, callers and callees, or obtain compiler-backed source facts. Use before selecting correspondence attachments or following calls. Read-only navigation, not link editing or a whole-program effect analysis.

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 5 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use the existing CLI rather than guessing declaration IDs or scanning source spellings for semantic references.

    line 8
  2. Distinctive:

    Read the selected package's instructions and identify the concrete behavior and source members of interest.

    line 16
  3. Very specific:

    Prefer --regression-result <native-engine-result.json> when a suitable recorded compiler context exists.

    line 20
  4. Distinctive:

    Keep selected sources explicit.

    line 23
  5. Distinctive:

    Use the package's selected wrapper/distribution.

    line 29
  6. Distinctive:

    Use document-relative inputs from the selected package, resolved to CLI paths against the caller's directory.

    line 36
  7. Very specific:

    Start with summary: it retains symbol matches and utility-query details without the full fingerprint map.

    line 57
  8. Very specific:

    Follow emitted call targets with show/describe when a selected source body exists.

    line 76
  9. Very specific:

    When already inspecting model/C# pairs, avoid a second lookup for every helper: use links inspect --section calls --callee-source true …

    line 79

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Navigate C# with compiler-backed queries

Use the existing CLI rather than guessing declaration IDs or scanning source spellings for semantic references. See Finding C# symbols for the authoritative command contract and agent safeguards. For model/C# comparison use correspondence inspection; for requested connection edits use link maintenance.

Establish the compilation first

  1. Read the selected package's instructions and identify the concrete behavior and source members of interest. Keep implementation sources explicit.
  2. Supply the actual references, defines, C# dialect, nullable and framework settings. Do not silently use host defaults for a different product context.
  3. Prefer --regression-result <native-engine-result.json> when a suitable recorded compiler context exists. For project regression stages use the receipt under <stage>/engine/result.json, not the enclosing stage receipt. Keep selected sources explicit; every selected file must have been observed by that compiler. Do not combine this option with context overrides or reconstruct its reference list in scripts.
  4. Report missing context or compiler diagnostics. Do not turn incomplete resolution into plausible partial answers. Analysis must not provision tools.

Use the package's selected wrapper/distribution. If it rejects a documented symbols action, treat that as a tool-capability/setup gap, not absence of the requested declaration. Pause for authorized documented setup; do not silently substitute a workspace build. After a tool change, obtain fresh applicable compiler evidence if the existing receipt fails validation.

The following independently synthetic examples run from the repository root. Use document-relative inputs from the selected package, resolved to CLI paths against the caller's directory; do not assume the tool changes that directory.

Choose the smallest query

QuestionQuery
What IDs correspond to this exact name?symbols find --name <name>; inspect all overloads
What declaration does this exact ID identify?symbols show --id <emitted-ID>
What declaration contains this source position?symbols resolve --file <selected-file> --line <line> --column <column>
Where is this symbol used in selected sources?symbols references --id <emitted-ID>
Which explicit calls target this method?symbols callers --id <emitted-ID>
What does this source member explicitly call?symbols callees --id <emitted-ID>
Can I read the member and its structural facts together?symbols describe --id <emitted-ID>
.\TlaLink.ps1 symbols find --source .\examples\MethodSummaries\src\PermitGate.cs --name TryTake --format summary
.\TlaLink.ps1 symbols describe --source .\examples\MethodSummaries\src\PermitGate.cs --id 'M:SummaryExample.PermitGate.TryTake' --format summary
.\TlaLink.ps1 symbols callees --source .\examples\MethodSummaries\src\PermitGate.cs --id 'M:SummaryExample.PermitGate.TryTake' --format summary

Start with summary: it retains symbol matches and utility-query details without the full fingerprint map. Request full JSON or a new --output file only when the complete pinned result is needed. Symbol commands do not need --annotation-version; annotation-aware inspect-source does.

Follow evidence, not names

  • Copy returned declaration IDs. Do not synthesize overload/generic IDs. In PowerShell, single-quote IDs containing backticks or pass emitted IDs as variables. Renames and signature changes can change IDs.
  • find --name matches the exact ordinal, case-sensitive simple name; it is not prefix, substring, wildcard or regex search. A miss for a name fragment does not establish that the intended declaration is absent. If the name is unknown, use resolve at a known declaration position or inspect the scoped source inventory, then reuse the emitted ID.
  • find returns candidates, not an automatic choice. resolve identifies a containing declaration, not the symbol referenced at the cursor. Positions are physical one-based lines and UTF-16 columns; multiple sources require --file. Do not use nearest-line guesses to clear ambiguity.
  • Follow emitted call targets with show/describe when a selected source body exists. references/callers can also target metadata IDs but still search only selected source; they do not scan framework bodies.
  • When already inspecting model/C# pairs, avoid a second lookup for every helper: use links inspect --section calls --callee-source true --callee-limit 5 --callee-max-lines 80 with that inspection's explicit context. Review page-local callee references, dispatch and omitted/truncated status first. Expand only a justified next source member with symbols describe. One-level callee excerpts are not recursive call/effect analysis and do not authorize crossing a declared interface boundary.
  • Use source inventory when the question needs member-wide access/await/cleanup facts or CFG flow, rather than replacing Roslyn analysis with regexes.

Interpret and report

Exit 0 means discovery/resolution succeeded, 1 means not found or an ambiguous position, and 2 means invalid input or incomplete compiler context. A successfully resolved target with zero scoped references is not unused globally. State the source selection, compiler context, IDs and unresolved questions in the answer.

Calls retain declared dispatch, not concrete interface/virtual/delegate or reflection behavior. Method groups, nameof and crefs are not call edges. Implicit accessors, disposal, awaiters and operator/runtime calls need separate review. Nested-function sites are marked, not assumed executed. Follow effects beyond selected sources when the claim needs them, while respecting explicit interface contracts. Resolved IDs and structural facts do not prove call effects, atomicity, concurrency correctness or source/model equivalence.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.